Expert Systems; Planning in AI; Robotics & AI

Expert Systems

What is an Expert System?

An Expert System is a computer program that uses AI to simulate the judgment and behavior of a domain expert.

Components

+----------------------------------------------+
|              EXPERT SYSTEM                   |
|                                              |
|    User Interface                            |
|         |                                    |
|    Inference Engine (Reasoning)              |
|         |                                    |
|    Knowledge Base (Facts + Rules)            |
|         |                                    |
|    Knowledge Acquisition Module             |
|    (from domain experts)                    |
+----------------------------------------------+

Reasoning Methods

Famous Expert Systems

Planning in AI

Planning is the process of finding a sequence of actions that leads from an initial state to a goal state.

Planning Representation — STRIPS

• Preconditions — what must be true before an action can execute
• Add list — facts that become true after the action
• Delete list — facts that become false after the action

Example: Robot Moving a Block

Action: PICK-UP(block)
  Preconditions: ON-TABLE(block), CLEAR(block), ARM-EMPTY
  Add: HOLDING(block)
  Delete: ON-TABLE(block), CLEAR(block), ARM-EMPTY

Planning Algorithms

Robotics & AI

AI in Robotics

AI enables robots to:

• Perceive — cameras, LIDAR, ultrasonic sensors
• Plan — path planning, task planning
• Act — motors, grippers, actuators
• Learn — adapt to new environments

Types of Robots

Key Robotics Technologies

Expert-System Reasoning and Planning

An expert system stores domain knowledge separately from the inference mechanism. Facts describe the current case while rules express conditions and conclusions. Forward chaining begins with facts and repeatedly fires applicable rules. Backward chaining begins with a goal and searches for rules whose conclusions can establish it. Conflict resolution selects among several applicable rules. An explanation facility records which rules and facts supported a conclusion.

Planning represents an initial state, a goal and actions with preconditions and effects. State-space search may progress forward from the initial state or backward from the goal. A plan must order actions so every precondition is true when required. Heuristics estimate remaining cost. Uncertain or changing environments require replanning or policies that choose actions from observed states.

Robotic Perception and Control

A robot combines sensors, actuators, computation and a control loop. Perception estimates relevant properties of the robot and environment from cameras, range sensors, encoders or inertial sensors. Localization estimates pose while mapping represents surroundings. Path planning finds a collision-free route and motion control converts it into actuator commands.

Feedback compares observed state with the desired state and corrects error. Sensor noise, delayed observations, actuator limits and moving obstacles make physical AI different from an ideal search problem. Autonomous robots therefore combine estimation, planning, control and safety constraints rather than relying on a single prediction model.