COMP 514 - Applied Robotics

• This was an Applied Robotics course that I took during my Masters at McGill. Some of the topics we covered are listed below.
  • Rigid-body Pose - Configuration space, task space, translations, rotations, reference frames, Euler angles, quaternions
  • Rigid-body Twists - Angular velocity, homogeneous transformations, rigid-body kinematics, twist, adjoint map
  • Trajectory Generation - Time scaling, paths, straight-line trajectories, first-order polynomial trajectories, third-order polynomial trajectories, fifth-order polynomial trajectories, trajectory generation via waypoints and issues
  • Potential Field - Attractive potential fields, repulsive potential fields, total potential fields, deadlocks, narrow passages, local minima, hessian, oscillation
  • Kinematics - Forward kinematics, velocity kinematics, jacobian, inverse kinematics, pseudo-inverse, redundancy, singularity, manipulability
  • Dynamics - Newton's second law, moment of inertia, angular momentum, newton-euler method, lagrangian formulation, rigid-body dynamics, rigid-body chains, inverse dynamic control, PD controller, stiffness, damping, redundancy
  • Force Control and Contacts - Constraint kinematics, dynamics with contact, hybrid motion-force control, compliance, impedance control
  • Collision Avoidance - Collision detection, collision distance, point clouds, bounding volumes, Gilbert-Johnson-Keerthi (GJK) algorithm for collision detection, probabilistic roadmap, bug algorithms, optimization vs sampling-based approaches
  • Constrained Optimization - Convex optimization, Newton's method, linear equality constraints, interior-point (barrier) method, inverse kinematics with redundancy resolution, singularity avoidance
  • Grasping - Object kinematics, object wrench, grasp matrix, form closure, force closure, internal force, maintaining contacts, Coulombs friction model, friction cone, grasping force optimization, where to grasp
  • Pick and Place - Reach object, grasp object, move object, release object, switching between subtasks, when to compensate object dynamics
  • Sensors - monocular and stereo cameras, lidar, radar, sonar, building maps from sensor data
  • Locomotion - Gait, bipedal gait, duty factor, stride, quadruped gait, underactuated robots, contact constraints, center-of-mass, floating-based representations, stance vs swing
  • Whole-body Control - Prioritized control, hierarchical kinematic/dynamic control, control as an optimization problem, base planning, base control, static and dynamic balance, support polygon, center of pressure
• We also had eight assignments throughout the course. They were all based on ROS1 in C++
  1. Husky Teleop Controller - Controlled the husky robot with keyboard commands
  2. Drone trajectory planning - Implemented a third-order polynomial trajectory to guide the drone to the target
  3. Inverse Kinematics - Implemented an inverse kinematics controller with redundancy resolution to move a simulated Kinova arm to a desired end effector pose
  4. Inverse Dynamics - Implemented an inverse dynamics controller with redundancy resolution to move a simulated Kinova arm to a desired end effector pose
  5. Orientation Control - Added end effector orientation control to my previous inverse dynamics controller
  6. Kinova arm bowling - Used ROS-actions to sequentially move the Kinova arm to a series of locations to knock down targets
  7. Pick-and-Place - Implemented a pick-and-place process to grasp an object and drop it in a bowl. See below for a working demo of my implementation!
  8. Collision avoidance - Implemented a navigation system to drive the husky robot to the goal, using lidar for collision avoidance