Agentic AI and Autonomous Decision-Making

Scope

Architectures and algorithms for autonomous goal-directed agents: planning, tool use, memory, reflection, LLM/foundation-model-driven agents, decision-making under uncertainty, and the transition from language reasoning to grounded action.

Research topics

Topics include, but are not limited to:

• Agent architectures (ReAct-style, hierarchical, neuro-symbolic)
• Long-horizon task planning
• Tool and API orchestration
• Memory and context management for persistent agents
• Grounding language instructions in actuation
• Reinforcement learning for sequential decision-making
• Hierarchical and options-based control
• Agent evaluation benchmarks
• Cost- and latency-aware agent design
• Human-in-the-loop approval workflows

Emerging challenges

• Compounding error over long horizons
• Verification of agent plans before physical execution
• Hallucination with actuation consequences
• Sample efficiency in the real world
• Reward specification
• Auditability of autonomous decisions
• Graceful degradation when a tool or sensor fails

Industry applications

• Autonomous maintenance scheduling
• Agentic process automation in manufacturing MES
• Warehouse task allocation
• Field-service dispatch
• Autonomous test and QA agents

Keywords

• agentic AI
• autonomous decision-making
• LLM agents
• task planning
• tool use
• reinforcement learning
• neuro-symbolic reasoning
• grounded language
• human-in-the-loop