Research
Projects

Imagine a future where small robotic teams inspect our infrastructure—bridges, pipelines, buildings, and satellites—finding leaks and cracks promptly. To conduct effective inspections, however, the robots must first know precisely where they are on these structures, a problem known as localization. This task consumes a significant portion of a robot's computational resources. Our project delves into how collaboration among robots can streamline this process, freeing up more computational resources for inspection tasks. Drawing inspiration from nature's collaborative strategies, such as animals taking turns watching for predators, we introduce a novel mechanism where robots self-organize into dedicated localizers or inspectors, optimizing group inspection productivity. What's remarkable is that this sacrifice, with some robots bearing the computational burden of localization for the group, is entirely self-organized. The robots autonomously decide their roles, adapting to changing environments seamlessly. This approach, validated through theoretical models, simulations, and hardware experiments, showcases effective inspection in dynamic environments, highlighting the potential of autonomous, self-organized robots in making our infrastructure safer.

​

Check out my Princeton Research Day Talk. 

Read the full paper: DARS2024Ramshanker.pdf (princeton.edu)

We have built a unique swarm of metal climbing robots that serve as a test bed for exploring different collective behavior algorithms for inspection purposes. The swarm includes 15 Rovables: mini 3-cm wide robots that can traverse metallic structures from any direction. Additionally, we have designed a larger metal climbing robot powered by the NVIDIA Jetson Nano, which is currently being used to generate SLAM-based maps of 3D manifolds. Our collective-localization algorithms have been successfully tested using the Rovable swarm.

We are modeling anti-predator vigilance in social animals. There is a fundamental trade-off when it comes to vigilance: animals do not get to eat when they look for predators. It has been observed in nature that this cost of vigilance is shared amongst the group in social animals but how vigilance/foraging strategies are negotiated and organized is poorly understood. We are currently developing theoretical models to explain this behavior based on the insights we have found in modeling cooperative robotic swarms.