INTELLIGENT AGENTS AND RATIONALITY FIT3080
• Agents and environments
Copyright By PowCoder代写 加微信 powcoder
• Rationality
• PEAS (Performance measure, Environment, Actuators, Sensors) • Environment types
• An agent is anything that can be viewed as perceiving its environment through sensors, and acting upon that environment through actuators
• Human agent:
eyes, ears and other organs for sensors
hands, legs, mouth and other body parts for actuators
• Robotic agent:
cameras and infrared range finders for sensors various motors for actuators
• Software / virtual agent:
keyboard input, file input, receiving from network screen, file output, sending to network
Agents and Environments
• The agent function f maps from percept histories P⋆ to actions A: f: P⋆ → A
• The agent program runs on the physical architecture to produce f • agent = architecture + program
Example: Vacuum-cleaner World and Agent
• Percepts: location and contents, e.g., [A, Dirty] or [B, Clean]
• Actions: Left, Right, Vacuum
• Program:
if status = Dirty return Vacuum else if Location = A return Right else if Location = B return Left
Rationality and Rational Agents
• Rationality depends on Performance measure
The agent’s prior knowledge of the environment The actions that the agent can perform
The percept sequence to date
• Definition:
For each possible percept sequence, a rational agent should select an action that
is expected to maximize its performance measure, given the evidence provided by the percept sequence and the agent’s built-in knowledge
Rational, Autonomous Agents
• Rationality is NOT omniscience
• Agents can perform actions to modify future percepts in order to obtain useful information
→ exploration, learning
• An agent is autonomous if its behavior is determined by its own experience
Task Environment – PEAS
To design a rational agent, we must specify the Task Environment • PEAS
Performance measure Environment
Actuators
PEAS – Automated Taxi example
• Performance measure
Safe, fast, legal, comfortable trip, minimize fuel consumption, maximize profit
• Environment
Road types, road contents, customers, operating conditions
• Actuators
Control over the car, interfaces for informing other vehicles and informing passengers
Cameras, sonar, speedometer, GPS, odometer, engine sensors, interface for receiving information from other vehicles and passengers (e.g., speech recognizer)
Performance measure considerations
Rational agent:
• maximize expected performance measure, considering both now and the future.
Performance Safe
Legal Comfortable
Fuel consumption
Every crash: -100 Arrives unharmed: +500 Every second -0.1
Every violation: -500 Every speed bump -10 Sudden turn, braking -50 Every liter: -0.5
General rule: better to design performance measures according to what one actually wants, than according to how one thinks an agent should behave
PEAS Example: Vision-Language Navigation
Chen et al., Topological Planning with Transformers for Vision-and-Language Navigation, CVPR 2021.
PEAS Example: Vision-Language Navigation
In break-out rooms,
• Check https://bringmeaspoon.org/ and the other links in this web-page (5 min) • Discuss what you think PEAS should be in this problem (5 min)
• Afterwards, we will discuss the responses
Environment Types (I)
The environment type largely determines the agent design
• Fully (partially) observable – An agent’s sensors give it access to the complete state of the environment at all times
• Known (unknown) – An agent knows the “laws” of the environment
• Single (multi) agent – An agent operating by itself in an environment
• Deterministic (stochastic) – The next state is completely determined by the current state and the action executed by the agent
Environment Types (II)
• Episodic (sequential) – The agent’s experience is divided into atomic episodes. The next episode does NOT depend on previous actions
In each episode an agent perceives a percept and performs a single action
• Static (dynamic) – The environment is unchanged while an agent is deliberating
• Discrete (continuous) – Pertains to number of states, the way time is handled, and number of percepts and actions
E.g., a state may be continuous, but actions may be discrete
Next week: Search algorithms
程序代写 CS代考 加微信: powcoder QQ: 1823890830 Email: powcoder@163.com