AI Agents and Multi-Agent Systems
How autonomous agents perceive, plan, learn, use tools and cooperate: agents and environments, search, classical planning, decisions under uncertainty and MDPs, reinforcement learning, language model agents and tool use, memory and retrieval, long-horizon reliability, game theory, task allocation and negotiation, multi-agent LLM systems, and evaluating and controlling agents.
AI-470 Year 4 3 credits 12 lessons · 4 h 54 min
Offered by HSM University · Artificial Intelligence
This is the Canadian version of this course. See the American version.
What you'll learn
- Define agents, rationality and PEAS, classify environments and compare reflex, model-based, goal-based, utility-based and learning agents
- Formulate search problems and apply breadth-first, uniform-cost, A*, local and adversarial search
- Represent planning problems with STRIPS and PDDL, use planning heuristics and HTNs, and handle execution failures with monitoring and replanning
- Apply expected utility, model sequential decisions as MDPs and POMDPs, and solve them with value iteration
- Explain reinforcement learning, exploration, Q-learning, deep RL and RLHF, and recognize reward hacking
- Build language model agents with the ReAct loop, well-designed tools, MCP integrations and appropriate workflow patterns
- Design agent memory with retrieval-augmented generation, context compaction, note-taking and permission-aware retrieval
- Improve long-horizon reliability with planning, verification, retries, grounded reflection, search and checkpoints
- Analyse multi-agent interactions with game theory, Nash equilibria, repeated games and mechanism design
- Allocate tasks with assignment algorithms, auctions and the contract net, and design safe negotiating agents
- Choose multi-agent LLM architectures, coordinate agents and evaluate when voting and debate help
- Evaluate agents rigorously and control them with least privilege, sandboxing, approvals, injection defences, monitoring and governance
Course content
12 lessons · 4 h 54 min total · final exam
- 1 What makes a system an agent 25 min · quiz
- 2 Search: how goal-based agents find their way 24 min · quiz
- 3 Classical planning: actions, preconditions and plans 25 min · quiz
- 4 Decisions under uncertainty: utility and Markov decision processes 23 min · quiz
- 5 Learning agents: reinforcement learning 25 min · quiz
- 6 Language model agents: reasoning, acting and tool use 25 min · quiz
- 7 Memory, retrieval and context engineering 25 min · quiz
- 8 Planning, reflection and reliability in language model agents 24 min · quiz
- 9 Multi-agent systems and game theory 25 min · quiz
- 10 Task allocation, negotiation and auctions 25 min · quiz
- 11 Multi-agent systems built from language models 25 min · quiz
- 12 Evaluating and controlling agents, and capstone 23 min · quiz
- E Final exam 20 questions
Requirements
Recommended background. You can start any course at any time.
Description
How autonomous agents perceive, plan, learn, use tools and cooperate: agents and environments, search, classical planning, decisions under uncertainty and MDPs, reinforcement learning, language model agents and tool use, memory and retrieval, long-horizon reliability, game theory, task allocation and negotiation, multi-agent LLM systems, and evaluating and controlling agents.
Part of the Science stream, typically taken in year 4.