Behaviour-based Robotics: Intelligent Control for Robots

Question 1

What are the different paradigms for robot control architectures?

Answer 1

• Deliberative - “Think (hard), then act”
• Reactive - “Don’t think, just (re)act”
• Hybrid - “Think and act concurrently (and separately)”
• Behavior-based - “Think the way you act”

Question 2

What is deliberative/reactive/hybrid control? What are the advantages/disadvantages?

Answer 2

Deliberative:
- Think hard and then act without refering to the sensors again.
- Robot performs a sequence of sense-plan-act steps.
- Pros: Can select best plan IFF it has enough time and the world model is acurate. Capable of prediction and learning (i.e., improve performance over time).
- Cons: Needs a lot of computation power. It’s slow. Thinking too long can be counterproductive. Dynamic environment => frequent replanning

Reactive:
- Don’t think, just react (Braitenberg vehicles)
- Actions are based on changes in the environment.
- Sensors are directly linked to the actuators
- Pros: Minimal computation => fast, real-time response. Does not require a world model = good for dynamic environments.
- Cons: No high-level commands. Difficult to define goals. Sensor noise/errors => poor performance.

Hybrid:
- Combination of deliberative and reactive control
- Think hard, than act with a ‘reactive loop’ (local planner that reacts to changes in a dynamic enviornment)
- Pros: Fast, dynamic response to changes in world, near-optimal and globally efficient.
- Cons: Difficult to coordinate reactive and deliberative parts (different time scales, different world representations) => Difficult to desig an effective mediater.
- Example: Mobile robot with a path planner (path from A to B) and a local planner that reacts to changes in the dynamic enviornment (e.g. obstacle avoidance)

Question 3

What is behaviour-based control? Can you give an example of a behaviour-based control architecture?

Answer 3

• Bottom-up structure
• Example: Sub-sumption:
  ○ Each behavior is placed in a hierarchy, with the lowest level behaviours in the bottom.
  ○ Behavior Input: sensors, other behaviours
  ○ Behavior Output: motor commands
  ○ Action selection: Behaviour with highest priority controls the motor

Question 4

What are the main characteristics of behaviour-based AI?

Answer 4

Situatedness: Being placed in an environment (Coupling between sensor values and actions).
Embodiment: Agent is placed in the environment. Behavior-based only considers how it is interacting with the enviornment
Intelligence and emergence: Intelligence comes from interaction with environment, individual behaviours are not intelligent by themselves

Question 5

What are the different strategies for co-ordinating behaviours in behaviour-based control?

Answer 5

• Competitive: “Winner take all” (e.g. subsumption, , Meas Behaviour network)
• Coorperrative: Combine actions from several behaviours to produce an action to send to the actuators (E.g., Motor Schemas)

Question 6

How do you design and build behaviour-based control architectures?

Answer 6

Behaviour architecture has different abstraction levels, from simple to complex. The behaviours are added incrementally (simple to complex), after every newly implemented behaviour the robot is tested.
Each behaviour can have inputs from sensors and other behaviours, and send outputs to actuators and other behaviours. Behaviours are simple, time-extended processes, not atomic actions.