Robot Intelligence Kevin Warwick Robot Intelligence Characteristics There

Robot Intelligence Kevin Warwick

Robot Intelligence Characteristics There are many different animals/insects: • Each has its own intelligence which is appropriate for that creature e. g. humans or spiders. • Each has its own range of sensors and actuators. • Each has its own usual environment – its intelligence works well in that environment. • In other environments a creature may well not be able to survive or prosper due to its intelligence and/or physical capabilities. • A creature’s intelligence may well not fit at all with another creature’s physical or mental requirements e. g. spiders playing tennis or humans flying around caves using inbuilt echo location. If a robot welder is stupid because it cannot make a cup of tea then so too is a human stupid because they do not have an ultra-violet sense.

Task Characteristics What sort of Job does the robot do? Inhospitable or Remote Environments: • Dangerous Environments (Chemical or Nuclear) e. g. Bomb Disposal/Mine Clearance/Military. • Environments too costly to send a human e. g. Undersea. • Environments that would take too long to send a human e. g. Space Exploration • Environments that are physically impossible to place a human e. g. Microscopic Environments Tasks with high fatigue factors and demanding duty cycles e. g. spot welding cars Tasks that are unpleasant for people to perform e. g. hoovering, cleaning, washing up.

Robot Hardware What makes a robot Any robot is made from a collection of various hardware components: - • Locomotion – how the robot moves within the environment. E. g. • • • Wheels or Legs. Sensing – how the robot obtains information about itself and the current state of the environment. E. g. Camera or Ultra-Sonic Range Finder. Reasoning – how the robot utilises the information obtained from its sensors to form decisions and actions. E. g. Reactive Controller or Computer running AI program. Communication – how the robot communicates to a human or machine operator and vice versa. E. g. Remote Control, Text Interface or Video Link.

Environment and Task Driven Constraints Different Locomotion Strategies must be chosen according to the task the robot has to perform and the particular environment the task must be performed in. • How fast must the robot be able to move? • Is the terrain rough or is it smooth like a floor? • Is the environment unstructured or is it well known and mapped?

Different Categories of Robot Terrestrial – robots that move on the ground. • Mostly wheeled but also with tracks or legs. • More exotic designs include climbing, rolling and slithering robots e. g. Robot Snakes. Aquatic – robots that operate in water. • Most are neutrally buoyant and use water jets or propellers to move. Airborne – robots that fly. • Typically helicopters or fixed wing aircraft. • Unusual types include dirigibles (air balloons) and controlled parachutes. Space – robots that operate in microgravity. • Independently propelled robots “free flyers” e. g. Satellites

Types of Control • Teleoperation – the robot is effectively remote • • controlled with images from a camera mounted on the robot being relayed back to the operator. e. g. Underwater ROVs, Bomb disposal robots. Semi-Autonomous – some decisions are made automatically by the robot but the overall behaviour is dictated remotely, such as interesting targets to explore. e. g. Mars Rover. Autonomous – all decisions are made by the robot itself. e. g. Dwarf Robot.

Locomotion (noun) Power of motion from place to place For a robot to be able to interact with its environment it must be able to: • Move within the environment in some manner • Sense the environment it moves through Kinematics: • The study of motion ignoring the forces that actually generate that motion. • Given some control inputs, how will a robot move? (FORWARD KINEMATICS) • Given some desired motion, which control inputs should be chosen in order to obtain the desired motion? (INVERSE KINEMATICS)

Statics and Static Stability • Forces and moments propagate through a robot’s structure • Static Stability – defines a set of gaits that allow a robot to remain “standing” without any control wheels/legs. Dynamics and Dynamic Stability • How forces generate accelerations that produce motion • Dynamic stability – having to dynamically control gait in order to ensure a robot does not fall over. e. g. a single legged hopping robot.

Robot Motion The motion of a robot will depend on the mechanism through which the motion is generated and supported • Wheeled • Legged • Aquatic • Flying • Rocket Propelled Let us consider, as an example, a description of these tasks for wheeled robots

Wheeled Mobile Robots • Wheels utilise friction • and ground contact to enable motion Lets consider the case of an ideal wheel pictured on the left.

The Ideal Wheel Consider an Ideal Wheel • Wheel rotates about the x-axis. • Motion is solely in the y-direction. • Measurement of wheel motion (odometry) from e. g. a wheel encoder is perfectly accurate. - A distance of 2. π. r in the y-direction is covered for every rotation of the wheel. Of course things aren’t quite that easy!

An actual wheel is considerably more complicated than the ideal: • May be lateral slip if there is insufficient traction • Rough terrain and bumps, compression and cohesion between the wheel and ground surfaces often leads to a loss in accuracy • Some of the resultant motion will be in the z direction

• Consequently, Odometry will be inaccurate: • Driven wheels are more prone to error due to the forces acting on the wheel • One technique – for measurement - is to use an additional non-driven, non-load bearing, light wheel to more accurately recover the motion of the robot – in terms of both distance covered and in direction if the light wheel has a castor. This can be used for as a good approximate for low velocity motion.

The Instantaneous Centre of Curvature [ICC]

The Instantaneous Centre of Curvature [ICC] • Consider the case where several wheels are in • • • contact with the surface – see previous slide. If all wheels in contact with the surface are to roll – then the axes of each wheel must intersect through a single centre of rotation – the ICC (case a). If no consistent ICC exists then the wheels cannot roll (case b). Not only must the ICC exist but each wheel’s velocity must be consistent with a rigid rotation about the ICC. E. g. If a set of three wheels were equidistant from the ICC they would all have to move at the same velocity.

Pose of a Robot and Frames of Reference • Vehicle on a plane has three degrees of freedom – two in translation (x, y) • Based on a fixed frame of reference, a {W} or World frame – one in orientation • = 0 is defined to point along the Wx direction • positive rotations are counter-clockwise – triplet (x, y, ) is defined as the pose of the robot

Types of Robot Motion • Holonomic Robots – can move in any direction instantaneously. This is clearly impossible as any real robot will have mass. • Omni-Directional Robots – can in practice move in any direction but takes time as the robot has mass. – Such robots are, in general, treated as being Holonomic. – E. g. Differential Drive with a castor wheel – dwarf robots. • Non-Holonomic Robots – this type of robot is limited in the way it can move e. g. Car Parking. – Normally limited by dynamic or kinematic restraints – e. g. limited turning ability

Pose of an ideal “holonomic” differential drive robot.

Pose of an ideal “holonomic” differential drive robot. • Consider the differential drive robot shown in the previous Slide. This is an ideal differential drive robot (with supporting castor wheel ignored for the purpose of simplicity) • Inter-wheel spacing is D. • Orientation of the robot is θ w. r. t. {W}. • Position of the robot is (x, y) w. r. t. {W}. • Where will the ICC be located?

Where is the ICC? • The ICC will be located at some point along the axis of the wheels. – If Vleft=Vright • Robot will move in a straight line forward/backwards – no change in • orientation i. e. ω=0. The ICC is effectively at ∞ therefore the radius of curvature is also ∞ (i. e. a straight line!). – If Vleft=-Vright • The robot will turn on the spot – no translation purely change in • orientation. The ICC is in the centre of the wheels and the radius of curvature is 0 (i. e. it rotates!). – If │Vleft│≠│Vright│ • The robot will both turn and move (i. e. both translation and rotation).

Where is the ICC? • We can start by writing some equations using the relationship between Angular and Linear Velocity.

Where is the ICC? • So for each wheel on the robot • [1] • [2] • Assuming we know VLeft and VRight (i. e. we can control the speed of the wheels!) • We can then solve the simultaneous equations for ω and R.

Solving for R: • From [1] • So: • Therefore: and from [2]

Solving for ω: • From [1] • Therefore: - from [2]

Robot Intelligence • Intellectual abilities/performance need to be • • compatible with actuators/sensors and must take account of mechanical reality. Can operate on equations of motion etc (normal route) Can operate on learned behavioural route (highly non-linear/human approach) Can operate on taught approach (Asimo) – programmed Can operate under remote control