1 CE801: Intelligent Systems and Robotics Lecture 3: Actuators and Localisation Prof. Dr. Hani Hagras
Robot Locomotion Robots might want to move in water, in the air, on land, in space.. 2 Most of the mechanisms for locomotion has equivalent in nature One important exception: the powered wheel! Biological energy storage and the muscular and hydraulic activation systems much better than man-made systems
Wheeled Locomotion: Car-like drive Car-like drive Two motors: one to drive, one to steer. 3
Wheeled Locomotion: Differential Drive 4 Two motors, one per wheel Wheels run at equal speeds for straight-line motion Wheels run at equal and opposite speeds to turn on the spot Other combinations of speeds lead to motion in a circular arc
Synchro drive One motor for translation, One motor for rotation Wheels mechanically coupled Turns without turning body For example Nomad200 robot 5
Skid-steer Drive More common for outdoor platforms Wheels or tracks Turn by applying different speed to wheels Skidding makes it hard to predict motion Extremely energy inefficient when friction is high 6
Exotic Wheeled Robots Segway platform with dynamic balance gives good height with small footprint and high acceleration Mars Rover has wheels on stalks to tackle large obstacles A robot with several omni-directional wheels on axles at angles can travel in any instantaneous direction (holonomic) 7
Non-holonomic constraint A vehicle is holonomic if the number of local degrees of freedom of movement equals the number of global degrees of freedom. A car is non-holonomic: the global degrees of freedom are motion in x,y and heading. However locally, a car can only move forward or turn. It cannot slide sideways. 8
Legged Locomotion Need at least 2 DOF to move leg forward Lift Swing Gait = sequence of lift and release events for the individual legs Robot legs often have 3 joints A fourth ankle joint can improve walking More DOFs increase complexity, weight and power requirements Human leg has more than 7 DOFs plus toes 9
Need to hop to move One Leg Robots Could in principle handle quite rough terrain Balance is the major challenge Static stability not possible The Raibert hopper from MIT Springs can capture kinetic energy and help increase efficiency 10
Biped Walking Biped walking can be approximated by a rolling polygon Smaller steps gives closer to rolling circle 11
Four Legged Robots (Quadrupeds) Can be statically stable when standing still Walking still challenging Example Sony Aibo, BigDog 12
Electronic actuators Hydraulic actuator Pneumatic actuators Air Muscle Actuators Shadow Hand by the Shadow Robot Company, London Actuated using air muscles (most joints have an opposite pair) 13
Electric Actuators 14
DC Motors Electric Actuators Most common motors, available in all sizes and types Simple control with voltage or pulse width modulation Advanced control with encoder and feedback Step Motors 15 Motion in precise increments controlled by electrical pulses Used when open-loop precise is required and desired torque is small (printers, disk drives)
Disadvantages Electric Actuators Inherently high speed with low torque, hence gear trains or power transmission units are needed 16 Gear backlash limits precision Electrical arcing may be a consideration in flammable atmospheres Problems of overheating in stalled condition Brakes are needed to lock them in position
Localisation 17 Autonomous robots are equipped with various sensors to see" the environment where they are committed to operate. One of the fundamental problems for autonomous operations is to find its location information. For ground mobile robots, the information includes a 2D coordinate (x; y) and an orientation (heading). Often the location information is called robot pose. For flying robots and underwater robots, a robot pose consists of 3D coordinate (x; y; z) and three angles around three axes. The localization problem, well known as a question Where am I?" in robotics, is stated as given the map of the environment and sensor's readings, to find out the pose of the robot.
Localisation 18 For outdoor environments, although the GPS can provide the global coordination, the accuracy is questionable for some tasks. For indoor environments, the GPS signal is not accurate or even not available. The on-board sensor based approaches are a key technique for localization, which is targeted by robotic research community for many years.
Triangulation and Trilateration The straightforward way for localization is the triangulation or trilateration methods in geometry. The triangulation is the process of determining the location of a point by measuring angles to it from known points. The trilateration is the process of determining the location of a point by measuring distances to it from known points. For 2D cases, the location of a point is calculated when the angle (or distance) measurements from three different beacons are known. 19
Trilateration 20
Triangulation 21
Measurement Noise 22 The localization results of the triangulation or trilateration methods are very sensitive to the noise in angle or distance measurements. The localization performance is not ideal for robot navigation. Mobile phone uses the trilateration method to provide the locating service. It does not measure the distance, but RSS (received signal strength) which is converted into distance value. Even worse, the three angle or distance measurements are not obtained simultaneously as the robot moves forward. The time compensation in the triangulation or trilateration methods has been developed, but the performance is still not satisfactory. Handling the noise in angle/distance measurements or any other sensory measurements leads to various probability based approaches. Kalman filter is a known probability based approach to deal with noise
Kalman Filter Kalman filter is a powerful tool in dealing with noise in the sensory measurements. 23 Kalman filter for robot localization has been successfully developed for many years The Kalman filter is an efficient recursive filter that estimates the state of a dynamic system from a series of noisy measurements.
KF Conceptual Overview To simplify the presentation, a one dimensional case is chosen to illustrate how to use KF for localization. The robot true position is x k at time step k, which we don't know and we want to estimate. The robot moves along the x direction with a dynamic state equation: 24 w k is a Gaussian noise with zero mean and variance q, representing the movement uncertainty. Assume the robot has a GPS device and it provides the noisy measurement z k at time step k. The measurement equation is: v k is a Gaussian noise with zero mean and variance r, representing the measurement noise.