Introduction to Mobile Robotics

Transcription

1 Introduction to Mobile Robotics Olivier Aycard Associate Professor University of Grenoble Laboratoire d Informatique de Grenoble olivier. 1/22

2 What is a robot? Robot = mechatronic system with perception, decision and action skills, capable of carrying out different tasks in the real world, in an autonomous way. Sensors Perception Decision Action Actuators 2/22

3 The robot of the day sensors sensors actuators actuators 1 raspberry pi3 Ubuntu + ROS Sensors 2 laserscanners Actuators 2 wheels driven by 2 motors + encoders 1 PC Ubuntu + ROS In charge of sensor data acquisition, processing & visualization; In charge of controlling actuators. 3/22

4 Sensor (1/2) A sensor is an instrument measuring a physical property of the environment; Sensors are imprecise and limited; The environment of a robot is generally complex, changing, unpredictable and uncertain; Understanding the world in which a robot evolves remains a challenge. Courtesy of sick 4/22

5 Sensor (2/2) Robair is equipped with a 2D laser scanner; The laser scanner has: a range of about 5.5 meters; a field of view of 240 degrees; an angular resolution of 1/3 degrees. The laser scanner costs about 1200 euros. Output: a table with 720 elements (r, Ɵ) Quality of data depends on distance, angle Polar to cartesian: X = r cos (Ɵ); Y = r sin (Ɵ). 5/22

6 Actuators of a mobile robot An actuator is a component of a machine that is responsible for moving or controlling a mechanism or system; An actuator controls a degree of freedom (rotation, translation); Actuators could be complex. 6/22

7 Actuators of robair (1/3) Robair has 2 wheels controlled by 2 motors; Robair is a differential drive robot; Simplest and most used kinematic model of robot. ω V R A monocycle describes a virtual circle of radius R; We have V = R ω 7/22

8 Actuators of robair (2/3) Robair has 2 wheels controlled by 2 motors; ω V r V w V l R We have V = R ω We have V r = (R + w/2) ω We have V l = (R w/2) ω We find: R= W 2 V l +V r V r V l 8/22

9 Actuators of robair (3/3) Finally, we have the direct kinematic model: ω = V r V l w (1) V= V l +V r 2 (2) Controlling (V l, V r ), we can determine (V, ω) But it is easier and more intuitive to control (V, ω) and determine (V l, V r ) (inverse kinematic model) V r = V + w 2 ω V l = V w 2 ω To simplify the control (at the beginning), we will perform translation OR rotation in place 9/22

10 y Estimation of motion: encoder While Robair is moving in its environment, we would like to know its position in this environment; Its position is determined by its position (x, y) in the environment + its orientation θ: (x, y, θ) θ x On each wheel, there is a system (named encoder) able to estimate the distance traveled by each wheel over a short time Δt 10/22

11 Estimation of position: odometry We call d l and d r the distance traveled by each wheel over Δt; d = d r +d l 2 using (2) Where d is the distance traveled and θ is the angle traveled x t = x t 1 + d cos θ y t = y t 1 + d sin(θ) θ t = θ t 1 + θ θ = d r d l w using (1) This is an estimation: with time the error associated to this estimation increases Drift problem 11/22

12 What is a robot? Robot = mechatronic system with perception, decision and action skills, capable of carrying out different tasks in the real world, in an autonomous way. Robot = artificial machine Body Robot Perception: Sensors Sensors Décision: Brain Computer Action: Members Actuators Autonomy: capacity to understand the current situation and to react in an approprious way taking into account the tasks to carry out. Sensors Perception Decision Action Actuators 12/22

13 Perception Perception Measurements Z X Vehicle State PERCEPTION M Static Objects Vehicle Motion Measurements U O Moving Objects Present Focus: interpretation of raw and noisy sensor data Identify static and dynamic part of sensor data Modeling static part of the environment Simultaneous Localization And Mapping (SLAM) Modeling dynamic part of the environment Detection And Tracking of Moving Objects (DATMO) 13/22

14 Decision/Plan of future actions Most of the time, a mobile robot has to move in its environment: It needs to plan its future actions The mobile robot has a map and it knows where it is in the map (position A for instance); It should reach an other position in the map (position B for instance) Question: How to get there? Answer: sequence of actions to go from A to B that is feasible and without collision. north West, South, East (12 times), North (6 times), West (twice) 14/22

15 Action/control/navigation The mobile robot has a sequence of actions to execute, to reach its goal: it has to execute this sequence of actions (session 3); Monitoring of execution: we monitor what happen and react if needed. We need to be able to estimate actions/motions of the mobile robot; Collision detection/avoidance: the mobile robot should be able to detect and avoid collision. 15/22

16 Organization of the course 11 sessions: 1. 4 sessions to introduce basic concepts: 1. Introduction to mobile robotics and IoT + introduction to perception (DATMO); 2. Lab on ROS + perception(datmo); 3. Lab on perception (DATMO) + introduction to decision and action; 4. Lab on decision and action + tests. Design and develop a follow me behavior 5. Tutorial on perception (Localization) + Tutorial on IoT 2. 7 sessions to apply and «get in depth» these basic concepts on a practical project. 16/22

17 Projects Design, develop and test software(/middleware) for a mobile robot application; Customization of projects and robots is possible Projects will be done by groups of 4 students MAX Have your own laptop, install ROS (under ubuntu) and do the tutorial on ROS Evaluation: 2 intermediate evaluations (week 5 and week 8-9) 1 final defense + demo 17/22

18 Examples of applications(1/3) Advanced driver assistant system (ADAS) or autonomous vehicles Darpa Urban Challenge 2007 Google car 2010 IP Prevent 2008 Google car /22

19 Examples of applications(2/3) Service robotics Roomba Robomow Dyia One Baxter Staubli 19/22

20 Examples of applications(3/3) Companion robots Paro Aibo Buddy Nao Pepper 20/22

21 Summary (1/2) A mobile robot is equipped with 2 kind of sensors: Exteroceptive sensors that give information about the environment (ie, laser scanner); Proprioceptive sensors that give information about the internal state of the robot (ie, odometer); A mobile robot is equipped with some actuators characterized by their degree of freedom; Robair is a differential drive robot; Sensors and actuators are imprecise and limited; The environment of a robot is generally complex, changing, unpredictable and uncertain. 21/22

22 Summary (2/2) The link between sensors and actuators is done in 3 steps: Sensors Perception Decision Action Actuators A mobile robot control architecture is: a Finite State Automaton; An instantiation of the «perception-decision-action» architecture Sensors Perception Decision Action Actuators Sensors Detect (and track) a moving person Decide actions to move to the moving person Move to the moving person Actuators 22/22