lec 1: course introduction artificial intelligence...
Post on 05-Sep-2020
2 Views
Preview:
TRANSCRIPT
Artificial Intelligence Techniques
for Mobile Robots
Teacher: Alessandro Sa�ottiRoom: T-2224Email: alessandro.saffiotti@oru.se
Lab assistant: Ali Abdul KhaliqRoom: T-2229Email: ali-abdul.khaliq@oru.se
Course home page:http://aass.oru.se/
~
asaffio/Teaching/AIMR/
Lec 1: Course Introduction
1. Course objectives
2. Mobile robots
– the “robot” word
– types of mobile robots
– inside a mobile robot
3. AI and mobile robots
– the first AI robot
– today’s AI robots
– limitations and risks
4. Autonomous robot navigation
– the goal
– the problems
5. Course organization
1c� A. Sa�otti 2018
Course Objectives
• Introduce some concepts, principles and techniquesof artificial intelligence (AI)
• Understand how these can be applied to physicalsystems, that are coupled to the environment viasensors and actuators
• Hands-on experience
– you will program your own robot that can move safely to
a goal position in an unknown environment
• Applies to any autonomous physical system
– camera-based monitoring and surveillance
– car safety and driver’s assistance
– autonomous vacuum cleaners
– smart phones
– . . . and so on!
2c� A. Sa�otti 2018
2. Mobile Robots
3c� A. Sa�otti 2018
The ‘robot’ word
• Invented by Karel Capek in his play “R.U.R.” (1921)(excerpt at http://pimacc.pima.edu/
~
gmcmillan/rur.html)
– Rossum’s Universal Robots
– ‘Robota’ = ‘forced labor’ in Czech
“Those who think to master the industry are mas-tered by it [...] The product of the human brainhas escaped the control of human hands. This isthe comedy of science.”
• Reused by Isaac Asimov in “Runaround” (1942)– 1950: he stated the three laws of robotics
1. A robot may not injure a human being, or, through in-action, allow a human being to come to harm.
2. A robot must obey orders given it by human beings,except where that would conflict with the first law.
3. A robot must protect its own existence except wherethat would conflict with the first or second law.
4c� A. Sa�otti 2018
Airborne robots
• “Predator”
– used by US army for autonomous recognition missions
– thousands UAVs for recognition and “weapon delivery”
– important discussions ongoing about ethics . . .
5c� A. Sa�otti 2018
Underwater robots
MBARI experimental underwater vehicle for autonomousexploration
6c� A. Sa�otti 2018
Tracked robots
Semi-autonomous robot for inspection and repair ofthe Chernobyl shelter (the sarcophage)
Tele-operated robot for inspection of pipelines
7c� A. Sa�otti 2018
Legged robots
Dante: CMU legged robot forsemi-autonomous volcano explo-ration
ASIMO: Honda humanoid robot
AIBO and QRIO:Sony entertainementrobots
8c� A. Sa�otti 2018
Wheeled robots
Autonomouswarehousemanagement
Autonomous cleaning
Brain-controlled wheelchairs
9c� A. Sa�otti 2018
Things you may find in a robot
• Actuation:
– motorized wheels (mobility)
– motorized joints (manipulation)
– speakers (interaction)
• Sensing:
– encoders on the motor axes (proprioception)
– inertial systems (proprioception)
– infrared sensors (exteroception)
– RGB cameras (exteroception)
– laser range-finders (exteroception)
– microphones (interaction)
• Computing:
– microcontrollers
– PCs
10c� A. Sa�otti 2018
Incremental optical encoders
• Measure incremental motion since last reading
• Advantages
– cheap and simple
– used in most mobile platforms
• Problems
– cannot tell the direction of rotation
– loose precision if some readings are missed
11c� A. Sa�otti 2018
Exteroceptive sensors: infrared
• Distance measurement:
– emit IR beam and measure amount of reflected light
– may be used to detect nearby obstacles
• Passive light measurement:
– may be used to detect a light source
• Problem
– reflected light strongly depends on surface
12c� A. Sa�otti 2018
Exteroceptive sensors: laser ranger
• Rotating laser beam
– time-of-flight
– phase shift
• Problems
– disturbance to the environment
– only senses on a given plane
– sensitive to environment light
– sensitive to object’s color
13c� A. Sa�otti 2018
Exteroceptive sensors: ToF camera
• Simultaneous luminosity and distance at each pixel
– illumination by array of modulated LEDs
• Problems
– disturbance to the environment
– sensitive to environment light
– short range
14c� A. Sa�otti 2018
Exteroceptive sensors: Kinect
• Simultaneous luminosity and distance at each pixel
– structured light from IR laser
• Fast posture detection software
– and SDK for robotic applications
• Problems
– only indoor
– short range
15c� A. Sa�otti 2018
3. AI and Mobile Robots
16c� A. Sa�otti 2018
AI in robotics: the early times
• “Shakey” the robot
• Stanford Artificial Intelligence Center, 1966
• First general purpose mobile robot
17c� A. Sa�otti 2018
AI in robotics: today
• Elderly assistance (ORU)
• Self-driving car (Google)
18c� A. Sa�otti 2018
AI in robotics: risks
19c� A. Sa�otti 2018
AI in robotics: risks
20c� A. Sa�otti 2018
4. Autonomous Robot Navigation
21c� A. Sa�otti 2018
The goal of navigation
• To reach a given location P
• Examples:
– Go to (x = 100, y = 200, ✓ = 90�)
– Go to room T2224
– Go to Oscar Wilde’s house
– Go to Place de la Bastille in Paris
– Go to a good observation position
• Possible ways to complicate the problem:
– Go to P in shortest time (optimal control)
– Go to P with least energy (optimal control)
– Go to P with max speed 1m/s (constraints)
– Be at P at 4:12 pm (deadlines)
22c� A. Sa�otti 2018
Facets of the navigation problem
• Need a map of the environment
• Make a navigation plan using this map
• Execute the plan
– move in a stable and safe way
– keep track of your position in the map
– detect and avoid obstacles and dangers
– notice exceptional situations and modify the plan
• All this needs the use of sensors
23c� A. Sa�otti 2018
Environment map
• Must include topological information
– which door to use to go from here to there?
• Must include geometric information
– how many meter to travel before turning left?
• Problem: find “right” level of detail
– if it is too abstract ) useless
– if it is too detailed ) unstable
24c� A. Sa�otti 2018
Planning
• Find a trajectory in the map that:
– goes from the start position to the goal position
– is collision-free
– is feasible given the robot’s kinematics and dynamics
– satisfies the extra constraints
• Problem: uncertainty
– in real environments, the configuration of the space may
not be fully known in advance, and it may change afterwards
25c� A. Sa�otti 2018
Execution
• Follow the planned trajectory
– guarantee physical stability
– keep track of your position in the map
• React to unexpected events
– use sensors to detect obstacles
– use sensors to detect failures in the plan
• Problem: sensor interpretation
– sensor data is noisy and di�cult to interpret correctly
26c� A. Sa�otti 2018
Re-planning
• Detect major discrepancies from the plan
– the plan is not feasible any more, or
– there is a new better oppourtunity
• Modify the plan
• Problem: when to re-plan?
– we want to react quickly to any new situation, but we do
not want to change our mind all the time
27c� A. Sa�otti 2018
What we shall see
• How to keep track of the robot’s position
• How plan a trajectory to a goal
• How to follow that trajectory
• How to detect and avoid obstacles
Note: not in this order, but. . .
28c� A. Sa�otti 2018
5. Course Organization
29c� A. Sa�otti 2018
Course Outline
• Lecture 1: course overview
– Lab 1: familiarize with the platforms
• Lecture 2: position estimation (by Ali)
– Lab 2: position estimation
• Lecture 3: closed-loop control
– Lab 3: go to a goal position
• Lecture 4: fuzzy rule-based control
– Lab 4: rules for “go to” and “avoid osbtacles”
• Lecture 5: path planning
– Lab 5: path planning and path following
• Lecture 6: the SPA architecture
– Lab 6: putting everything together
• Lecture 7: Course wrap-up
– Lab 7: the final challenge
30c� A. Sa�otti 2018
Course Schedule
See course home page
Note: Attendace is compulsory!
31c� A. Sa�otti 2018
Laboratories
• On real mobile robots
– “table” robot
– programmed in C
• Plus simulated mobile robots
– same robot
– same program
• Groups of 3-4 people
– you’ll decide them at first lab
• Each lab has
– some basic tasks
– some optional tasks
• Brief report
– see instructions on web site
– due to Ali 10 days the lab has been assigned
– must be checked and approved by Ali
• Final lab: “The Final Challenge”
32c� A. Sa�otti 2018
The Final Challenge
• Replaces the written exam
• An extended laboratory work (in groups)
– put together the work of previous labs
– some basic tasks, plus several optional tasks
– you have about 2 weeks to do it
• Final report (individual)
– max 10 pages, in English or in Swedish
– discuss your solution: why and how
– attach your own code as appendix
• Evaluation (individual)
– demonstrate the working program to Ali (on Oct 25)
– give the report to me (by Oct 28)
– fix a date to discuss it with me (in week 44)
• Your final score will be based on:
– submitting all your lab reports correctly and in time
– number of options successfully performed
– correctness and quality of your final report
– face-to-face discussion of your final report
33c� A. Sa�otti 2018
top related