A physics-informed digital twin and onboard model predictive control for an autonomous active particle.
Conventional active particles move with fixed motility: they convert energy into motion, but they cannot sense where they are, predict where they are going, or decide what to do next. Closing that loop onboard — on a centimetre-scale robot with a real actuator and a real power budget — is the step that turns an active particle into an agent (Mammadli et al., 2026).
Approach
The work builds on a programmable bristlebot and proceeds in two stages. First, a physics-informed digital twin: a kinematic model fitted to measured trajectory statistics, which then generates long synthetic trajectories that remain statistically faithful to the real device. Second, that same kinematic model becomes the prediction step of a model predictive controller running on the robot, which is what makes autonomous trajectory tracking possible.
System identification calibrates the twin; the twin's kinematic model becomes the predictor inside the robot's onboard controller.
The loop matters in both directions. Measurements from the physical robot identify the model parameters, and the identified model is what the onboard controller uses to anticipate its own motion.
Result
The brainbot executes a non-trivial target path accurately, which is a proof of principle for a particle that senses its state, predicts its evolution, and computes control inputs onboard. Combining physical modeling, data-driven parameter identification, and control in one framework gives a platform for machine-learning-enabled multi-agent studies — the groundwork for intelligent, adaptive active matter.
Establishing adaptive particles that sense their state, anticipate their evolution, and compute control inputs onboard has been a major challenge in non-equilibrium physics. We address this challenge by realizing an autonomous brainbot, building on a recently developed programmable bristlebot. First, we construct a physics-informed digital twin of the device, based on a kinematic model that reproduces measured trajectory statistics and generates long, statistically faithful synthetic trajectories. The kinematics forms the foundation for implementing onboard model predictive control (MPC), enabling autonomous trajectory tracking, demonstrated by accurate execution of a non-trivial target path. This provides a proof of principle for a brainbot that senses its state, predicts its evolution, and computes control inputs onboard, unlike conventional active particles with fixed motility, thereby transforming the brainbot into an agentic physical entity. By integrating physical modeling, data-driven parameter identification, and control into a unified framework, our approach provides a scalable platform for machine-learning-enabled multi-agent studies and lays the groundwork for intelligent, adaptive active matter.
@article{mammadli2026brainbot,title={Physics-informed digital twin and onboard control of a brainbot for intelligent active matter},author={Mammadli, Isa and Shrestha, Prajol and Pande, Jayant and Novkoski, Filip and Mohapatra, Siddhant and Noirhomme, Martial and Maier, Andreas and Vandewalle, Nicolas and Smith, Ana-Sun{\v c}ana},journal={Physical Review Research},volume={8},number={3},pages={033164},year={2026},publisher={American Physical Society},doi={10.1103/rd1t-h2v6},note={My contribution: design and onboard implementation of the model predictive control (MPC) for autonomous trajectory tracking.}}
