DIRo Lab - Research

Research Directions

Current Research Directions

(For complete list of papers go to Publications)

Multi-Agent Reinforcement Learning for Multi-robot Systems in Warehouse Robotics

Safe RL for Autonomous Vehicles using Control Barrier Functions

Aerial Robotics for Precision Operations including Parcel Delivery, Autonomous Landing

Multi-robot Formation Control in Presence of Actuator Delays

Flexible Object Transport Using Ground Robot Teams

Prior Research Directions

Swarm Networks: My research in swarm networks aims to develop models to explain cohesive maneuvers observed in natural swarming systems, such as parallel turning maneuvers in starling flocks, and to use those models for control of engineered multi-agent networks in robotics and transportation.

Standard Models: Turning maneuver dissipates as it propagates from leader agent (red) to followers (such as in blue and green), leading to formation loss, using standard network control models.

Turning_maneuver_consensus.mp4

(Need to sign in to Google account to play the videos)

Proposed Approach: Turning maneuver propagates from leader agent (red) to followers (such as in blue and green) without distortion, leading to cohesive turning maneuver, using proposed delayed self reinforcement approach.

Turning_maneuver_stronglydamped.mp4

Example: Synchronized orientation response of swarm networks during turning maneuvers.

The central agent, leader in red, starts turning, and the maneuver information propagates through the network to other follower agents (such as in blue and red at two different distances from the leader agent).

Work done with: Santosh Devasia

Publications in swarm networks:

Anuj Tiwari, Santosh Devasia, James J Riley. Low distortion information propagation with noise suppression in swarm networks, Proceedings of National Academy of Sciences (PNAS) 2023 (PDF) https://doi.org/10.1073/pnas.221994812

Tiwari, Anuj, and Santosh Devasia. Improving network’s transition cohesion by approximating strongly damped waves using delayed self reinforcement. In 2021 Seventh Indian Control Conference (ICC). IEEE. (weblink)

Networked Robotics: My research in networked robotics focuses on improving cooperative operation of multiple robots for tasks such as transporting flexible objects, with applications in advanced composite manufacturing for aerospace industry, and automation in warehouse and delivery industry.

ACC_videosubmission_anuj.mp4

Work done with: Santosh Devasia

(Need to sign in to Google account to play the videos)

Communication-free Cohesive Flexible-Object Transport using Decentralized Robot Networks.mp4

Work done with: Yoshua Gombo, Santosh Devasia

Publications in networked robotics:

Anuj Tiwari, Santosh Devasia. Decentralized Cohesive Response During Transitions for Higher-Order Agents Under Network Delays, in IEEE Transactions on Automatic Control, vol. 67, no. 11, pp. 6303-6309, Nov. 2022, doi: 10.1109/TAC.2022.3183035. (web, PDF)

Y. Gombo, A. Tiwari and S. Devasia, Accelerated-Gradient-Based Flexible-Object Transport With Decentralized Robot Teams, in IEEE Robotics and Automation Letters, vol. 6, no. 1, pp. 151-158, Jan. 2021, doi: 10.1109/LRA.2020.3036569. (web, PDF)

A. Tiwari and S. Devasia, Rapid Transitions With Robust Accelerated Delayed-Self-Reinforcement for Consensus-Based Networks, in IEEE Transactions on Control Systems Technology, doi: 10.1109/TCST.2020.3032853. (weblink, arxiv)

Gombo, Yoshua, Tiwari, Anuj, and Santosh Devasia. Communication-free cohesive flexible-object transport using decentralized robot networks. American Control Conference, 2021. IEEE. (weblink)

Tiwari, Anuj, and Santosh Devasia. Cohesive Velocity Transitions in Robotic Platoons Using Nesterov-type Accelerated Delayed Self Reinforcement (A-DSR). In 2019 Sixth Indian Control Conference (ICC). IEEE. (weblink)

Connected and Autonomous Vehicles: My research in autonomous mobility develops cohesive network control techniques, which maintain vehicle formation during transitions, and not only in the steady state, can lead to improved inter-vehicle spacing control in connected vehicles.

Traffic_Intersection.mp4

Right: Standard spacing control

Left: Proposed decentralized spacing control

(Need to sign in to Google account to play the video)

Research direction - Efficient and safe operation of connected automated vehicles (CAVs) is essential, for instance, for fuel economy and emission reduction, and increased traffic throughput, using a platoon of closely spaced heavy-duty trucks. My research in autonomous mobility develops cohesive network control techniques, which maintain vehicle formation during transitions, and not only in the steady state, can lead to improved inter-vehicle spacing control in connected vehicles.

Example: Longitudinal spacing control and velocity tracking by improving velocity cohesion using delayed self reinforcement (DSR). Following is a simulated traffic intersection simulation where:

i) Right: Standard local spacing control methods lead to expansion of vehicle network and loss of capacity.

ii) Left: Proposed decentralized spacing control using delayed self reinforcement which leads to cohesive velocity changes among vehicles during speed transitions, and increased intersection capacity.

Work done with: Yudong Lin, Santosh Devasia

Publications in connected vehicle networks:

Yudong Lin, Anuj Tiwari, Brian Fabien, Santosh Devasia. Constant Spacing Connected Platoons with Robustness to Communication Delays, IEEE Transactions on Intelligent Transportation Systems, doi: 10.1109/TITS.2022.3224635 (web).

Tiwari, Anuj, and Santosh Devasia. Safely increasing capacity of traffic intersections with mixed autonomous vehicles using delayed self reinforcement. In 2022 Seventh Indian Control Conference (ICC). IEEE. (PDF)

Contact at anujt@smail.iitm.ac.in

Google Sites

Report abuse