DOC PREVIEW
Berkeley COMPSCI 287 - Lecture 1: Introduction

This preview shows page 1 out of 4 pages.

Save
View full document
View full document
Premium Document
Do you want full access? Go Premium and unlock all 4 pages.
Access to all documents
Download any document
Ad free experience
Premium Document
Do you want full access? Go Premium and unlock all 4 pages.
Access to all documents
Download any document
Ad free experience

Unformatted text preview:

Page 1CS 287: Advanced RoboticsFall 2009Lecture 1: IntroductionPieter AbbeelUC Berkeley EECS http://www.cs.berkeley.edu/~pabbeel/cs287-fa09 Instructor: Pieter Abbeel Lectures: Tuesdays and Thursdays, 12:30pm-2:00pm, 405 Soda Hall Office Hours: Thursdays 2:00-3:00pm, and by email arrangement. In 746 Sutardja Dai Hallwww Communication: Announcements: webpage Email: [email protected] Office hours: Thursday 2-3pm + by email arrangement, 746 SDH  Enrollment: Undergrads stay after lecture and see meAnnouncements Prerequisites: Familiarity with mathematical proofs, probability, algorithms, linear algebra, calculus. Ability to implement algorithmic ideas in code. Strong interest in robotics Work and grading Four large assignments (4 * 15%) One smaller assignment (5%) Open-ended final project (35%) Collaboration policy: Students may discuss assignments with each other. However, each student must code up their solutions independently and write down their answers independently. Class Details Learn the issues and techniques underneath state of the art robotic systems Build and experiment with some of the prevalent algorithms Be able to understand research papers in the field Main conferences: ICRA, IROS, RSS, ISER, ISRR Main journals: IJRR, T-RO, Autonomous Robots Try out some ideas / extensions of your ownClass Goals Logistics --- questions? [textbook slide forthcoming] A few sample robotic success stories Outline of topics to be coveredLecture outlinePage 2 Darpa Grand Challenge First long-distance driverless car competition 2004: CMU vehicle drove 7.36 out of 150 miles 2005: 5 teams finished, Stanford team won Darpa Urban Challenge (2007) Urban environment: other vehicles present 6 teams finished (CMU won) Ernst Dickmanns / Mercedes Benz: autonomous car on European highways Human in car for interventions Paris highway and 1758km trip Munich -> Odense, lane changes at up to 140km/h; longest autonomous stretch: 158kmDriverless carsKalman filtering, Lyapunov, LQR, mapping, (terrain & object recognition)Autonomous Helicopter Flight[Coates, Abbeel & Ng]Kalman filtering, model-predictive control, LQR, system ID, trajectory learningFour-legged locomotioninverse reinforcement learning, hierarchical RL, value iteration, receding horizon control, motion planning[Kolter, Abbeel & Ng]Two-legged locomotion[Tedrake +al.]TD learning, policy search, Poincare map, stabilityMapping“baseline” : Raw odometry data + laser range finder scans[Video from W. Burgard and D. Haehnel]MappingFastSLAM: particle filter + occupancy grid mapping[Video from W. Burgard and D. Haehnel]Page 3Mobile ManipulationSLAM, localization, motion planning for navigation and grasping, grasp point selection, (visual category recognition, speech recognition and synthesis)[Quigley, Gould, Saxena, Ng + al.] Control: underactuation, controllability, Lyapunov, dynamic programming, LQR, feedback linearization, MPC  Estimation: Bayes filters, KF, EKF, UKF, particle filter, occupancy grid mapping, EKF slam, GraphSLAM, SEIF, FastSLAM Manipulation and grasping: force closure, grasp point selection, visual servo-ing, more sub-topics tbd Reinforcement learning: value iteration, policy iteration, linear programming, Q learning, TD, value function approximation, Sarsa, LSTD, LSPI, policy gradient, inverse reinforcement learning, reward shaping, hierarchical reinforcement learning, inference based methods, exploration vs. exploitation  Brief coverage of: system identification, simulation, pomdps, k-armed bandits, separation principle  Case studies: autonomous helicopter, Darpa Grand/Urban Challenge, walking, mobile manipulation. Outline of Topics Overarching theme: mathematically capture What makes control problems hard What techniques do we have available to tackle the hard problems E.g.: “Helicopters have underactuated, non-minimum phase, highly non-linear and stochastic (within our modeling capabilities) dynamics.”  Hard or easy to control?1. Control Under-actuated vs. fully actuated Example: acrobot swing-up and balance task1. Control (ctd) Other mathematical formalizations of what makes some control problems easy/hard: Linear vs. non-linear Minimum-phase vs. non-minimum phase Deterministic vs. stochastic Solution and proof techniques we will study: Lyapunov, dynamic programming, LQR, feedback linearization, MPC1. Control (ctd) Bayes filters: KF, EKF, UKF, particle filter One of the key estimation problems in robotics: Simultaneous Localization And Mapping (SLAM) Essence: compute posterior over robot pose(s) and environment map given (i) Sensor model (ii) Robot motion model Challenge: Computationally impractical to compute exact posterior because this is a very high-dimensional distribution to represent [You will benefit from 281A for this part of the course.]2. EstimationPage 4 Extensive mathematical theory on grasping: force closure, types of contact, robustness of grasp Empirical studies showcasing the relatively small vocabulary of grasps being used by humans (compared to the number of degrees of freedom in the human hand) Perception: grasp point detection3. Grasping and Manipulation Learning to act, often in discrete state spaces value iteration, policy iteration, linear programming, Q learning, TD, value function approximation, Sarsa, LSTD, LSPI, policy gradient, inverse reinforcement learning, reward shaping, hierarchical reinforcement learning, inference based methods, exploration vs. exploitation4. Reinforcement learning system identification: frequency domain vs. time domain Simulation / FEM Pomdps k-armed bandits separation principle …5. Misc. Topics Control Tedrake lecture notes 6.832: https://svn.csail.mit.edu/russt_public/6.832/underactuated.pdf Estimation Probabilistic Robotics, Thrun, Burgard and Fox. Manipulation and grasping - Reinforcement learning Sutton and Barto, Reinforcement Learning (free online) Misc. topics -Reading materials Next lecture we will start with our study of


View Full Document
Download Lecture 1: Introduction
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Lecture 1: Introduction and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Lecture 1: Introduction 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?