Research @CoPheLab
Research in the CoPhe Lab spans three interconnected verticals: collective intelligence, male infertility, and traffic & pedestrian dynamics—unified by a focus on collectives and emergent phenomena. Our work is inherently interdisciplinary, bringing together students from diverse backgrounds to study complex systems and collective behaviour. While rooted in computation, we increasingly integrate experiments and field studies across all three verticals, bridging theory, data, and real-world systems.
Collective intelligence
swarm of drones, computational ethology, reinforcement learning
Collaborations
Prof Vishwesha Guttal, TEELAB, CES, IISc Bangalore
Prof Ganga Prasath, AMBE, IIT Madras
Prof Jason R Picardo, Chemical Engineering, IIT Bombay
Ongoing projects
How intruders learn to navigate dynamic obstacles?
Deriving the stochastic partial differential equations that govern a Vicsek flock.
Learning the rules of the game: a game-theoretic view of complex systems.
We study emergence—how simple local interactions give rise to complex, intelligent behavior at the collective level. From ant colonies that self-organize without central control, to schools of fish and flocks of birds that move as cohesive units, nature shows that intelligence can arise from interaction.
Yet, such behavior is hard to predict or design. Simple rules can produce order—like flocking or lane formation in pedestrian crowds—or failure, as in the tragedy of the commons. This unpredictability is a central challenge for building engineered collectives such as drone swarms and autonomous systems.
Our goal is to develop a predictive science of collective intelligence. We investigate how interactions, environment, and learning shape emergent behavior, and how agents adapt strategies within a collective using reinforcement learning. Complementing this, we use data-driven methods to uncover governing equations of collective dynamics, including stochastic PDE models.
Ultimately, we aim to move from observing collective intelligence to designing it.
Male infertility research
sperm cell motility, microfluidic mimics, hydrodynamics of swimming, active matter
Collaborations
Dr Mohan S Kamat, Reproductive medicine and surgery, CMC Vellore
Prof Kiran Raj M, AMBE, IIT Madras
Prof K V Venkatesh, Chemical engineering, IIT Bombay, METFLUX pvt ltd.
Ongoing projects
Understanding how interactions affect the inference of sperm cells?
Combined estimation of motility and morphology from microscopy videos.
Developing microfluidic devices that mimic the female reproductive system.
Role of hydrodynamics of the swimming sperm cells on the motility and collective migration.
Sperm cells do not operate in isolation—they navigate as part of a complex, interacting collective within the female reproductive system. Their success depends not just on individual motility, but on interactions with boundaries, fluid flows, other cells, and chemical and thermal cues.
Yet, current clinical standards largely ignore this complexity, relying on simplified, individual-level metrics. We aim to transform this paradigm by bringing a systems-level understanding of sperm behavior into diagnosis.
Our work bridges theory, data, and experimentation. We develop realistic models and digital twins of sperm collectives that capture non-idealities—variations in tail beating, morphology, and environmental noise—moving beyond overly idealized descriptions. In parallel, we design microfluidic platforms that replicate physiologically relevant conditions of the female reproductive tract, enabling us to study sperm behavior in environments that matter.
By linking collective dynamics to clinical observables, we seek to redefine diagnostic standards for male infertility—making them more predictive, integrative, and grounded in real biological function.
Traffic & pedestrian dynamics
indian crowds, av path planning, pedestrian-vehicle systems, decision making
Collaborations
Prof Lelitha Devi V, Civil Engineering (transportation), IIT Madras
Prof Bhargava C, Civil Engineering (transportation), IIT Madras
Prof Dorine Duveis, Department of Transport & Planning, TU Delft, Netherlands
Dr Yufei Yuan, Department of Transport & Planning, TU Delft, Netherlands
Ongoing projects
Autonomous vehicle path planning using a dynamic artificial potential field.
Indian pedestrian dynamics: effect of group cohesion on the fundamental diagrams.
Reinforcement learning (PPO) to optimally shepherd an intruder through a crowd.
Data-driven discovery of the social forces from pedestrian movement data.
Towards identifying strategies to increase ambulance mobility through dense crowds on Indian roads.
Indian traffic presents a rich, real-world laboratory for studying decision-making in complex collectives. It is defined by heterogeneity—from cycles to trucks—lane-less interactions, frequent rule-breaking, dense packing, and seemingly risky maneuvers with low margins for error. Yet, despite this apparent chaos, traffic often remains functional.
This complexity poses a fundamental challenge for autonomy: systems designed for structured environments struggle here. If autonomous vehicles do not think like Indian drivers, they will not succeed on Indian roads. Beyond detection and lane-following, they must anticipate intent, predict trajectories, and make context-aware decisions in real time. Our work uses AI and learning-based approaches to develop such bottom-up intelligence for navigation in unstructured, dynamic environments.
We extend this perspective to pedestrian crowds, where collective behavior is equally nuanced. Dense crowds can spontaneously reorganize—making way for an ambulance or adjusting flow without explicit coordination. We study how such large-scale, high-dimensional decision-making emerges, and how it can be modeled and predicted.
By combining data, theory, and learning, we aim to develop models of Indian traffic and pedestrian systems that inform the design of autonomous agents, guide infrastructure planning, and enable early prediction of critical events such as congestion and crowd disasters.