Karteikarten
361
Karten in diesem Stapel (149)
- 1What is the definition of artificial intelligence?The automation of activities that we associate with human thinking
- 2What is the general idea portrayed by searles chinese room?That behaviour is insufficient to assess intelligence
- 3How do we formulate a problem in basic search?State space - set of all states that are reachable from an initial state, by any sequence of actions
Path - sequence of actions leading from one state to another
Initial state - a known initial state
Operations - set of all possible actions - 4How do we formulate a goal in basic search?With a well defined function that identifies both the goal states AND the conditions on which it can reach the goal
e.g. Goal = save money
State space = flights + cost
Path = sequence of flights - 5What is the 8-queens problem?Problem placing 8 queens on an 8x8 chess board so that no two queens can threaten each other
- 6What are two problem formulations of the 8-queens problem?`1- place a queen in row 1 in any position, repeat this for each row but a different position each time
State space = 8!
2 - place a queen in any open space, repeat until all queens cannot threaten each other
State space = 64! - 7What four metrics do we use to measure search performance?Completeness
Time complexity
Space complexity
Optimality - 8What is completeness (search performance context)?Does the strategy theorised guarantee to find a solution, if one exists?
- 9What is time complexity (search performance context)?How long does the strategy take to find a solution?
- 10What is space complexity (search performance context)?How much memory does the strategy require to carry out the search?
- 11What is optimality (search performance context)?Does the strategy find the best quality solution if multiple solutions exist?
- 12What is breadth-first search, and how does it perform in terms of search performance metrics?Breadth-first search is a search algorithm which explores each node at each depth until it's found a solution (goes left to right)
This algorithm is complete
O(b^d) b= branching factor, d = depth (this is for time & space)
Optimal when the path cost is non-decreasing - 13What is uniform cost search, and how does it perform in terms of search performance metrics?Uniform cost search is a search algorithm based on breadth-first search, but each arc has an associated cost. The node expanded is always the one with the least cost
f(n)=g(n)
complete & optimal
time & space = O(b^c/e)
b = branching factor, c = cost of optimal solution, e = path/arc cost - 14What is depth-first search, and how does it perform in terms of search performance metrics?Depth-first search is an algorithm that explores child nodes (from left to right) to the goal/ to it's leaf nodes if no goal found, then backtracks to the most recent node
Not complete (think infinite depth), non-optimal (multiple goal nodes)
time = O(b^m) b = branching factor, m = maximum depth
space = O(bm) (doesn't keep all nodes in memory like breadth-first - 15How do we solve the 'infinite depth' problem in depth-first search?Add a maximum depth, and this produces depth-limited search. This still isn't optimal/complete as depth limit could be less than graph depth, and if the limit is greater than the depth then it's non-optimal
time & space = O(b^l), l = depth limit - 16What is iterative deepening, and how does it perform in terms of search performance metrics?Iterative deepening is an algorithm which explores each node at each possible depth, but recomputes all the nodes before when moving deeper (combination of breadth-first & depth-first)
Complete (b = infinite) & optimal (non-decreasing path cost)
time = O(b^d), b = branching factor, d = depth
space = O(bd) - 17What is bi-directional search, and how does it perform in terms of search performance metrics?The bi-directional search algorithm runs two searches simultaneously, one from the start node and one from the goal node, and the solution path is where the two searches meet
Complete (if using breath-first for both), optimal (non-decreasing cost)
time & space = O(b^d/2) - 18What is an intelligent agent?An autonomous entity which observes through sensors and acts up an environment using actuators and is directed towards achieving a goal
- 19What are the 5 variations of intelligent agents?Reflexive
Goal-based
Problem solving
Learning
Model - 20What is a reflexive agent?A reflexive agent is a type of intelligent agent which has a fixed set of condition-action rules e.g if-car-in front-is-breaking THEN apply-breaks
- 21What is a goal-based agent?A variant of intelligent agents which allows the agent to choose among multiple possibilities, selecting one which reaches a goal state
- 22What is a problem solving agent?A variant of intelligent agents which allow the agent to define how desirable a particular state is, enabling the finding of the sequence of actions which lead to the goal state
- 23How do we improve blind searching/avoid repeating states?Through the use of a heuristic function
A heuristic function produces an exact estimate of the evaluation function, which describes how valuable each state was in obtaining the goal.
Enabling the expansion of the nodes with the best value - 24What is best-first search?Best-first search is a search algorithm which makes use of a heuristic function to determine the best nodes to expand
f(n)=h(n) - 25What are the two types of best-first search (covered in this module)?Greedy search (aka greedy best-first search)
A* search - 26What is a greedy search, and how does it perform in terms of search performance metrics?A variation of best-first search in which a heuristic function is used to minimise the estimated cost to a goal
Non-optimal (due to local-maxima), non-complete (infinite depth)
time & space = O(b^m) b = branching factor, m = max depth - 27What is the A* search?A variation of best-first search and is a combination of greedy & uniform cost. Uses a heuristic function to minimise the total path cost, where the heuristic function is; f(n) = g(n) + h(n), f = total estimated cost of path through node n, g = cost so far to reach n, h = estimate cost from n to goal
- 28how does A* search perform in terms of search performance metrics?Complete - if there is a finite branching factor and if every operator cost is is greater than the branching factor
Optimal - if the heuristic function is admissible
Time & space = O(b^d), b = branching factor, d = depth - 29In respect to A* search, what are some complexities and what does a heuristic function need to be?Admissible(reasonable) - meaning that it never over-estimates the cost to the goal
Memory - all generated nodes need to be kept in memory
Computation - bad quality HF could cause space and time complexity to increase dramatically - 30What are the two types of memory bounded searches (in this module at least)?Iterative deepening depth first search
Iterative deepening A* search - 31How does iterative deepening depth-first search work?Perform depth-first search repeatedly with increasing depth limits, until the goal is found
- 32How does iterative deepening depth-first search reduce memory usage?Because the maximum nodes required to store is equal to the depth
- 33How does iterative deepening A* search work?Produce a depth-first search with a limit on f (total estimate cost of path through n) as opposed to depth
Each iteration, perform this ^ step. If f is greater than a threshold we cut off that branch (don't explore it)
Threshold to begin with = estimate of the cost at the initial state
At each iteration, the threshold used for the next iteration is the minimum cost of all values that exceeded the current threshold - 34What are the two types of problem solving methods (used in this module)?Optimisation - makes use of a function to maximise/minimise under constraints
Classification
Unsupervised - clustering
Supervised - decision trees, neural networks - 35What is a decision tree, and how do they work?A tree in which internal nodes represent a 'test' on an attribute, and each branch represents the outcome of the 'test'
Leaf nodes represent the values/outcome - 36What are the three types of nodes that 'typically' make up a decision tree?Decision/internal nodes - represented as squares
Chance nodes - represented as circles
End/leaf nodes - represented as triangles - 37How do we decide which attribute to split on in decision trees?We'd have to find out which one reduces the uncertainty, the attribute which reduces the uncertainty the most is the attribute we split on (entropy + IG)
- 38What is entropy?A measure of the amount of uncertainty in a probability distribution
-(X)Log2(X) - 39How is average entropy calculated?Overall probability distribution multiplied by the entropy
e.g. if we have 20/50, 10/50, 0/50
entropy = -(20/50)log2(20/50)-(10/50)log2(10/50)-0 = 0.99
average entropy = (30/50) * 0.99 = 0.595 - 40How is information gain calculated?Entropy before - average entropy
Best attribute = highest information gain - 41What is overfitting and noise?Noise - large amounts of meaningless information
Overfitting - error in which the learning system tightly fits the given training data so much that it would be inaccurate in predicting the outcomes of the untrained data - 42What is probability theory?Probability theory describes what an intelligent agent should believe, but on the basis of evidence
- 43What is utility theory?Utility theory describes what an intelligent agent wants
- 44What is decision theory?Decision theory combines utility & probability theory to describe what an intelligent agent should do
- 45What are the four basic probability variants?Unconditional - used for predicting e.g. P(event) = 0.23
Random variables - used to predict unpredictable things e.g. P(weather = rain) = 0.80
Probability distribution - used for things like graphs e.g. CDF
Conditional probability - used to predict something, given something else e.g. P(cavity | toothache) = 0.3 - 46What is Bayes rule?Theorem that describes the probability of an event, based on prior knowledge of conditions that might be related to the event
e.g. P(A | B) = P(B | A) P(A) / P(B) - 47In belief networks, what is the conditional independence property (CIP)?The probability of an event happening, based on a series of conditions
e.g. P(X | A, B, C)
what is the probability of X, given the probability of A, B and C - 48What are the steps involved in building a belief network?1 - Choose a set of relevant variables
2 - Choose an ordering for the variables
3 - While there are variables left, pick a variable (a1) and add a node to the network
4 - pick parents (a1) as the min set of nodes that satisfy the CIP
5 - define the CIP table for node a1 - 49What are the three types of probabilities?Frequentist - Probabilities which could from experiments e.g. '9/10 dentists agree'
Objectivist - Probabilities that are real aspects of the universe
Subjectivist - Probabilities that characterise beliefs e.g. 'in my opinion...' - 50What is a genetic algorithm?Genetic algorithm's are meta-heuristics, and what is meant by this is they are used to generate/select optimal heuristics
Based on Darwin's theory of evolution, in which the fittest individuals are selected for reproduction in order to produce offspring for the next generation - 51What are the steps involved in generating a genetic algorithm?Initial population
Fitness function
Selection
Crossover
Mutation
Accepting
Replace - 52What is the initial population (aka start) stage in genetic algorithms?Initial population - generate a random population of strings
A random string/individual is known as a chromosome
An individual is characterised by a set of parameters, these are called genes - 53What is the fitness function stage in genetic algorithms?Fitness function - determines how fit an individual is. A fitness score is given to each individual and the probability that an individual will be selected for reproduction is based on its fitness score
- 54What is the selection stage in genetic algorithms?Selection - choose the fittest individuals and let them pass their 'genes' onto the next generation
Two pairs of individuals (parents) are selected based on their fitness scores. - 55What is the crossover stage in genetic algorithms?Crossover - using a crossover probability/point, crossover parents to form new offspring. If no crossover is performed, the offspring will be an exact copy of the parents
- 56What is the mutation stage in genetic algorithms?Mutation - using a mutation probability, mutate new offspring at some position in the chromosome
- 57Why do we mutate chromosomes in genetic algorithms?To maintain diversity with a population and prevent finding a solution to early (could be a better one)
- 58What is the accepting stage in genetic algorithms?Accepting - the newly generated offspring are placed in a new generation
- 59What is the replace stage in genetic algorithms?Test if the end condition has been satisfied, if it has then stop and return the best solution ELSE
Replace - use the newly generated population to repeat the process, if an optimal solution hasn't been found - 60What are the four ways we can encode a chromosome?Binary encoding
Permutation encoding
Value encoding
Tree encoding - 61How does binary encoding work(genetic algorithms)?- Each chromosome is a binary string
- The meaning of the string can vary, but each bit is a characteristic of the solution
- Whole string can represent a number - 62How does permutation encoding work (genetic algorithms)?- Every chromosome is a string of numbers, which represent numbers in a sequence
e.g. travelling salesman (ordering problems) - 63How does value encoding work (genetic algorithms)?- Every chromosome is a string which is a value
- Value can be anything related to the problem - 64How does tree encoding work (genetic algorithms)?- Every chromosome is a tree of objects e.g. a function in a programming language
- 65How do we define the fitness amount a) when the correct answers is known and b) when it isn't?Correct answer known - fitness metric is devised toward the correct answer
Correct answer unknown - fitness metric is a estimate of the value of the solution - 66What are the four ways we select fit chromosomes?Easy way - take n most fit individuals
Roulette wheel selection
Rank selection
Elitism - 67What problems occur when we select the n most fit individuals when selecting chromosomes?Lose diversity - all succeeding chromosomes will be incredibly similar
Get stuck in local-minima - meaning we the circle of chromosomes we 'breed' stays small - 68What is roulette wheel selection?Allocate space based on the population of fitness scores
Still random, but chromosomes with higher fitness scores are more likely to be chosen (sort of like a bias)
Suffers problems when the fitness scores differ massively - 69What is rank selection?Works by giving each chromosome fitness scores based on a rank
e.g. worst = 1 fitness, best = n fitness
Can take a lot of time to find a solution as the best chromosomes won't differ much - 70What is elitism selection?Always keep the best solution we've discovered so far, just in case we don't find a better one (copy the current best into the succeeding populations)
- 71How does binary mutation in binary encoding work?In binary encoding, a binary mutation is just a bit inversion
Can prevent problems with local minima - 72How does binary crossover in binary encoding work?Switch parts of the encoding between two samples
e.g. 11001011 + 11011111
Cross over the first 5(C1) and the last 3(C2)
= 11001111 + 11011011
aka single point crossover - 73What are the four types of crossover in binary encoding?Single point crossover - a subsection of the first chromosome and a subsection of the second chromosome are swapped
Two point crossover - 2 subsections are swaped
Uniform crossover - 3 subsections
Arithmetic crossover - two samples are extracted and added together to create a new sample - 74How does permutation crossover in permutation encoding work?Crosses over a subsection numbers (uses single point crossover)
- 75How does permutation mutation work in permutation encoding?Two numbers are selected (one from each chromosome) and exchanged
- 76How does value crossover work in value encoding?Makes use of all the binary crossover techniques (single, two, uniform, arithmetic)
- 77How does value mutation work in value encoding, and how does it differ amongst different types?Real number - add/subtract a small value
Alphabetical characters - select a gene (letter) and replace it with what follows it in the alphabet
Directions - rotate it (north, west, east, south) - 78What are the two types of replacement strategies (genetic algorithms)?Steady state selection - in every generation, select a few high fitness chromosomes for creating new offspring, remove low fitness ones and replace with new offspring
Generational selection - large portion of the population is selected and crossed (typically half the individuals), the resulting offspring is mutated and inserted into the population, thus replacing the old individuals. - 79What are the advantages and disadvantages to using genetic algorithms?+ Parallelism
+ Easy to implement (oki)
+ Lower chances of local min/maxima probs
- Encoding/fitness can be difficult to choose
- Expensive (computationally)
- Strong dependence on params
- Not complete - 80What is a genotype?An encoding that is composed of a directed graph structure where each node =
Dimensions for a rigid part (robot)
Attachment joints
Joint types at those points
Maximum recursion depth - 81What is phenotype?An encoding that produces a hierarchy of 3D parts in a physics rich environment
- 82What are iterative improvement algorithms?Algorithms that make modifications to the solution at each iteration to improve the quality
- 83How do iterative improvement algorithms work?By checking if the state contains the solution, if it does then the path you take through the search tree is irrelevant
Start with a complete config and then make modifications improving quality - 84What is hill-climbing search and how does it work?An optimisation technique for solving computationally hard problems
Works by moving in the direction of increasing value within a local area of the state space
(Very efficient as doesn't need to maintain a tree) - 85What is the main problems that can occur with the hill-climbing search?Local extrema - occurs when the algorithm stops making progress towards an optimal solution, mainly due to the lack of immediate improvement of adjacent states
- 86What is minimax (game playing)?A decision rule used in decision theory for minimising the possible loss for a worst case scenario
- 87What is alpha-beta pruning?A variation of minimax
but sometimes we already know what a sub-tree will evaluate to without computing the value of every leaf node and filtering upwards
This is alpha-beta pruning
General rule : B <= A prune - 88What is expectiminimax?A variation of minimax, but when we encounter nodes which are determined by chance, compute the expected value based on the probability distribution
(By changing the evaluation function values, we change the outcome) - 89What is a path and what is a trajectory?Path = continuous curve connection between two samples/configs
Trajectory = Path that is parameterised by time - 90What is the configuration space (motion planning)?The space for all possible configurations
Represented with a Cartesian frame
Allows the treating of a robot as a single point - 91What are the four methods for motion planning?Probabilistic roadmap
Visibility graphs
Cell decomposition
Potential fields - 92What is the probabilistic roadmap method of motion planning?A method in which the configuration space is sampled at random
The samples are kept in 'free space'
Pairs of samples are they connected via simple paths - 93What are the issues with the probabilistic roadmap method of motion planningNot complete - might not complete task
Sampling strategy - due to being critical, slightly off changes a lot - 94What is the visibility graphs method of motion planning?A method of motion planning which consists of a non-directed graph with vertices (nodes) that can be the start, goal and obstacles
Edges (arcs) connect all of the vertices
suffers the 'graph search problems' meaning, nodes may have to be revisited - 95How do visibility graphs work?- If a path is not blocked by an obstacle, then the shortest path is a straight line
- If a path is blocked, the shortest path will be from one of the vertices
- Id 3-D if a path is blocked, then the shortest path might be from an edge - 96What are the four types of cell decomposition (motion planning)?Rasterization
Iterative
Strips
Cylindrical decomposition - 97What is the rasterization type of cell decomposition?- Select a pixel resolution
- Mark each pixel - blocked, free, partially blocked
- Only generate paths that go through free squares - 98What is the iterative type of cell decomposition?- For partially blocked squares, iteratively decompose them in to smaller units
(poses the question; when do you stop? this is an issue) - 99What is the strips type of cell decomposition?- Uses non-square cells of variable size
- Any cells that can share an edge are adjacent
- Robots can navigate between adjacent cells by moving from cell centre to the midpoint of a shared edge - 100What is the cylindrical decomposition type of cell decomposition?- Based upon critical points
- To find critical points sweep a line across c-space (?)
- Any point where the boundary curve is vertical marks a critical point - 101What is the potential fields type of cell decomposition?- Run a scalar function over free space
- Ideal field is smooth, with a global minimum at goal with no local minima
- Grows to infinity near obstacles - 102What is the issue when using potential fields (motion planning)?Computing an ideal potential field is likely to be as computationally expensive as path planning itself
- 103What are the types of ambiguity in NLP?Pragmatic - meaning in the context
Semantic - meaning of the sentence
Anaphoric - what/who does pronoun reference
Syntactic - structure/grammar
Lexical ambiguity - multiple meanings to a word
Metonym - one object stands for another e.g. suit -> business man
Metaphor - figure of speech in which a word/phrase is applied to an object/action - 104What are the three types of Part-Of-Speech (POS) tagging?Rule based POS tagging
Unigram probability POS tagging
Probabilistic POS tagging - 105What is rule based POS tagging?Rule based POS tagging - apply most likely tag for each word
- 106What is unigram probability POS tagging?Unigram probability POS tagging - Track how often a word appears in different contexts. Improves the result iteratively by applying a set of patching rules, using immediate context
- 107What is probabilistic POS tagging?Probabilistic POS tagging - if we have a sequence of words each with one or more potential tags, then we can choose the most likely sequence of tags by calculating the probability of all possible sequences of tags
Relies on the probability that one tag follows another & the probability of a word being assigned a tag from all possible tags - 108What two probabilities does the probabilistic POS tagger rely on?Relies on the probability that one tag follows another
Relies on the probability of a word being assigned a tag from all possible tags - 109What is the Hidden Markov Model (HMM)?Formal foundation for making probabilistic models of linear sequence - used for the 'labelling problem'
- 110What is first & second order HMM respectively?First order - estimates used for tag transition (before the current n-gram, and following the current n-gram) probabilities are derived from bi-grams of the proceeding/succeeding n-gram
Second order - Same of first, but uses trigrams (2 n-grams proceeding, 2 n-gram succeeding - 111What is the CLAWS (Constituent Likelihood Automatic Word tagging System) process?Tokenisation
POS tag assignment
Rule-based contextual disambiguation
^mixture of rule based & probabilistic - 112How many levels of annotation are there in NLP processing?6
- 113What is involved in level 1 of annotation (NLP Processing)?Orthography - how italics, initial capital letters, periods and delimitation of words by spaces are interpreted
Phonology - how syllables are interpreted
Phonetic/Phonemic - the defining of vowels and consonants
Morphological - how prefixes, suffixes and stems are interpreted - 114What is involved in level 2 of annotation ( NLP processing)?Lemma - dictionary headwords e.g. BE could be tagged 'is', 'are'
Prosody - transcription of stress, intonation and pulses - 115What is involved in level 3 of annotation ( NLP processing)?Grammar - assigning word-class labels for major parts of speech, sub-classes e.g. singular/plural nouns
- 116What is involved in level 4 of annotation ( NLP processing)?Syntactic - looking at where noun-phrases, verb phrases fit together
- 117What is involved in level 5 of annotation ( NLP processing)?Semantics - meaning of words
- 118What is involved in level 6 of annotation ( NLP processing)?Discoursal - description of how two segments are logically connected to each other
Pragmatics - how context contributes to the meaning - 119What are the seven layers of NLP?Pragmatics
Discourse
Semantic
Syntactic
Lexical
Morphological
Phonetic - 120What is meant by word sense?Meaning of a word
- 121What is meant by linguistics?The study of a language
- 122What is annotation/tagging?The marking up of linguistics information in text
- 123What is a semantic tagger?A system for carrying out automated semantic annotation, which is the process of attaching meaningful information to the appropriate linguistic units
- 124What are the four semantic study areas in NLP?Lexical semantics - study of meaning of linguistic lexical units e.g. single words
Pragmatic semantics - meaning of language in context e.g. bark
Grammatical semantics - study of the meaning of words which have relevance to grammatical role/syntax
Logical semantics - study of relations between NLP and formal logical systems - 125What are the four language meaning units in NLP?Morphological unit - smallest meaningful unit of language, including simple words and sub-words i.e. prefixes/anti/less
Lexical unit - single word, part of a word or word chain e.g. computer, take care
Syntactic unit - a word or group of words that functions as a unit within a sentence
Larger units - paragraphs, chapters - 126What is a multi-word expression (MWE)An important lexical unit which is an idiom or a certain set of words e.g. between a rock and a hard place
- 127What are the four causes of word sense ambiguity?Homophony - different words with the same sound e.g. new/knew
Homonym - different words have the same spelling e.g. read
Polysemy - a word may have more than one sense e.g. bark
Figure-of-speech - word/MWE used in a non-literal sense e.g. butterflies in my stomach - 128What are the five word sense relations?Synonym - similar
Antonym - opposite
Hypernym - think of it like a superclass in programming e.g. colour is a hypernym of red
Hyponym - think of it like a sub-class in programming e.g. red is hyponym of colour
Meroynym - part-of relation e.g. a wheel is a meronym of a car - 129Why do we semantically annotate?Automate information analysis and management
Allows the conducting of data analysis to greater extent
Allows the semantic analysis of various languages without the need for a system for each - 130What are three approaches used for word sense disambiguation?Rule-based - reply on certain rules to select the word sense
Statistical methods - measure probability of word senses given context features
Hybrid - combination of the previous methods, but depend on context information e.g words/n-grams proceeding/succeeding - 131What is machine learning?Machine learning is when a machine changes its structure, program or data based on its input, in such a manner that its expected future performance improves
- 132What are the limitations of machine learning?Data - not having the relevant data in order to build an AI
Ethics - preventing a learning bias
Discrimination - automatic decisions on real life issues - 133What is the learning process of an AI?Observe - provide the ai with lots of data and examples so it can learn
Finding patterns - the more data the more likely it'll find patterns within the data
Learning - after learning from the data the machine will be able to make better decisions - 134What is the primary aim of an artificial intelligence?To allow computers to learn (via training) automatically without human assistance, and be able to adjust their actions accordingly in order to make better decisions
- 135What is 'mutual information'?A measure using statistical association between candidate keys and context words
Calculates the probability that a word belongs in a context - 136What are the applications of AI?Speech recognition - alexa
face recognition - Tag a friend on FB
Fraud detection - paypal
Travel shortest paths - google maps
Recommendations - netflix - 137What is the goal of training (ai)?To allow the ai to learn from data with the hopes of finding relationships, make decisions and evaluate their confidence
Quality/quantity has a massive affect on the success of an AI - 138What is the goal of testing (ai)?To evaluate the trained model with data which the ai hasn't seen before
Usually done by spliting the training data into training and testing data (typically 90-10 or 70-30) - 139What is cross validation?A method of training the ai to prevent overfitting
Works by spliting the data into folds, for instance if we're training with 90% of the data, we could split the data into 9 folds and train the data in 9 different 90% - 140What is supervised machine learning?Teaching the machine in a way we can monitor
Works by applying what the machine as learned in the past to new data, using labelled examples to predict future events - 141What is unsupervised machine learning?Machine is given unlabelled training data to learn from, with the hopes it finds patterns
- 142Why do we use feature reduction?Remove redundant data
Reduces over-fitting
Improves accuracy
Reduces training time - 143What are the five things that determine how much training data we require?Task
Performance
Noise in the data
Noise in the features
Complexity of the model - 144What are features of text data?Verbs
Nouns
Pronouns
Adverbs
Function words
Word length
# of syllables
Topic
Word frequency - 145What are the inputs of the motion planning problem?Geometry of robots and obstacles
Kinematics of robots (degrees of freedom)
Start/goal position - 146What are the outputs of the motion planning problem?Continuous sequence of collision free robot configs from the start to the goal
- 147What is the goal of the motion planning problem?Move robot from start position to goal position, whilst avoiding obstacles
Compute collision-free path for a rigid articulated object among static obstacles - 148What is the basic idea behind celluar decomposition (motion planning)?The basic idea behind this method is that a path between the initial configuration and the goal configuration can be determined by subdividing the free space of the robot's configuration into smaller regions called cells
- 149What is CAPTCHA?Complete automated public turing test to tell computers and humans apart
Difficult for computers to recognise what is in image,
but easy for humans
Informationen
- Beschreibung
- Karteikarten am 361, erstellt von Mac Howarth am 08/05/2018.
- Erstellt von
- Mac Howarth
- Aufrufe
- 3
- Bewertung