EIES-Lab · The Hong Kong Polytechnic University

Laboratory for
Embodied Intelligence
in Energy Systems

Energy Shapes Intelligence. Intelligence Reshapes Energy Systems.

Our Mission

Understanding the Coupled Dynamics of Energy and Embodied Intelligence

Underlying Mechanism: Energy–Information–Action Coupling

We investigate how energy constraints and physical dynamics shape intelligent perception, decision-making, and autonomous adaptation, and how embodied intelligence, through physical actions and feedback, transforms the operation of energy systems.

Our Approach

Closing the Loop Between Energy Perception and Physical Action

We integrate energy-system perception, physics-grounded world models, and energy-constrained decision-making with physical action and feedback, enabling autonomous adaptation in complex and dynamic energy systems.

aPhysical energy environmentbEmbodied agenttt+1t+2Perception & state estimationMulti-source partial observationsState estimate with uncertaintyWorld modelState · dynamics · observationsPhysics, safety, consequencesDecision & policyPlanning, control & safe decisionsRevenue · cost · resilienceObservationsreal-time dataPhysical feedbackreward · cost · safetyActionRollout · planningPhysical embodied actionActing on the real energy systemActive sensingDevice controlEnergy dispatchRobot operation

aPhysical energy environment

bEmbodied agent

Perception & state estimation

Multi-source partial observations

State estimate with uncertainty

World model

State · dynamics · observations

Physics, safety, consequences

Rollout · planning

Decision & policy

Planning, control & safe decisions

Revenue · cost · resilience

Physical embodied action

Acting on the real energy system

  • Active sensing
  • Device control
  • Energy dispatch
  • Robot operation
a · The physical energy environment (solar radiation, a 3D city, PV, grid, building loads, storage, EVs and low-altitude drones) sends observations to b · the embodied agent. Perception estimates the system state; the world model predicts how the system evolves and what actions will cause; decision and policy plan through imagined rollouts and choose a safe action. Physical action (active sensing, device control, energy dispatch, robot operation) changes the environment, and physical feedback (reward, cost, safety) returns to the agent to update its world model and policy.

Our ResearchResearch Thrusts

Three connected steps of one loop. Read more about our research →

Energy-System Perception & Observability

Energy-System Perception & Observability

What information is sufficient for reliable energy decisions?

Physics-Grounded Energy World Models

Physics-Grounded Energy World Models

How do energy systems respond to changes and actions?

Energy-Constrained Embodied Decision & Control

Energy-Constrained Embodied Decision & Control

How can energy agents act safely and adaptively?

Flagship Research ScenariosEmbodied Intelligence in Energy Systems

A unified embodied intelligence framework across diverse energy systems, physical constraints, and feedback dynamics. Read more about our scenarios →

Autonomous Urban BIPV & Solar–Storage Systems
Active

Autonomous Urban BIPV & Solar–Storage Systems

Embodied Intelligence for Data Center Energy & Cooling
Active

Embodied Intelligence for Data Center Energy & Cooling

Climate-Resilient Urban Renewable Energy Networks
Active

Climate-Resilient Urban Renewable Energy Networks

LatestNews

All news →

[Placeholder] Paper accepted

A new paper on energy-aware embodied intelligence has been accepted. Replace with a real publication.

Join Us

Build the Future

Meet the team →   Contact us →

We welcome students, researchers, and collaborators with interests in energy systems science, GeoAI, physical AI, physics-based modeling, and real-world validation and deployment.