Research & Work
I design robotics and AI systems that can operate in crowded, dynamic, and infrastructure-constrained environments. My work spans perception, navigation, fleet management, and neurosymbolic AI — with a strong emphasis on resilience, explainability, and deployability at scale.
Academic Foundation
Master's Thesis: Beyond Rewards
Learning from Richer Supervision | IIT Madras, 2012
Explored how human instructions can serve as structured learning signals in RL. Introduced π-Instructions (action guidance) and Φ-Instructions (state abstraction), anticipating modern human-in-the-loop and advisor-style AI systems. Read more →
Industry Research
Alongside academic research, I have led and contributed to industry-driven robotics research focused on solving real-world deployment challenges. These works bridge theoretical ideas with systems that operate in production settings.
Fast Rescheduling for Multi-Agent Plan Execution in Dynamic Urban Environments
Pradyot Korupolu et al. • IEEE CASE, 2025
Reframed multi-robot plan execution as a machine scheduling problem. Proposed fast rescheduling methods to handle delays, failures, and environmental changes in large-scale urban robot fleets.
Context-Aware GAN-Based Image Retrieval for Coarse Localization of Autonomous Robots
Pradyot Korupolu et al. • IEEE/RSJ IROS, 2024
Introduced a GAN-based, context-aware image retrieval framework for coarse localization. Enabled localization in environments where traditional SLAM and GPS are unreliable.
A Low-Cost Bin Picking Solution for E-Commerce Warehouse Fulfillment Centers
Pradyot Korupolu et al. • ACRA, 2019
Designed a cost-effective robotic bin-picking system for large-scale warehouse operations. Demonstrated that reliable automation is achievable without high-end hardware.
Shock Reduction for Autonomous Navigation on Rough Terrain
Pradyot Korupolu et al. • Advances in Robotics (ACM), 2013
Proposed a Difference of Normals–based approach to reduce shocks in autonomous ground vehicles. Focused on improving hardware longevity and localization stability in real deployments.
Research Areas
Autonomous Delivery Robots
Leading the development and deployment of autonomous delivery robots across real-world environments at Ottonomy. Robots deployed globally for intralogistics, food and retail delivery, and airport operations.
VLL / Multi-LLM Advisory Architecture
An approach to keep LLMs as advisors, not drivers — combining vision-language models with symbolic rules for safe, auditable actions. This framework allows foundational models to guide decision-making without direct control.
Resilient AI-in-the-loop Systems
Frameworks for graceful degradation, human-in-the-loop recovery, and fleet-level monitoring. Building systems that can handle failures elegantly and maintain operational continuity in real-world deployments.
Neurosymbolic AI for Robotics
Combining neural networks with symbolic reasoning for explainable and customizable robot behavior. Using Semantic Rule Languages (SRLs) to maintain transparency and control in AI-driven systems.
Perception & Navigation
Developing robust perception systems for autonomous navigation in complex environments. Focus on multi-sensor fusion, dynamic obstacle avoidance, and real-time path planning.
Fleet Management
Scalable architectures for managing multiple autonomous robots simultaneously. Coordinating delivery schedules, optimizing routes, and handling fleet-wide anomalies.
Research Philosophy
My approach to robotics research emphasizes:
- Real-world Deployability: Technology that works outside the lab, in unpredictable environments
- Explainability: AI systems that can explain their decisions and actions
- Resilience: Graceful degradation and recovery from failures
- Scalability: Solutions that work for one robot or a thousand
- Safety First: Human-centric design with multiple layers of safety checks
Current Focus
Currently focused on:
- Scaling autonomous delivery across diverse environments (airports, campuses, urban areas)
- Integrating foundational models (LLMs/VLMs) as advisors in robotic decision-making
- Building infrastructure for AI-in-the-loop systems at scale
- Developing frameworks for explainable and auditable autonomous systems