Bibek Poudel

Bibek Poudel, Sai Swaminathan, Weizi Li

Designing a transit network requires many sequential route extension decisions, but their quality is often visible only after the full network is assembled. This delayed-feedback challenge lies at the heart of the Transit Route Network Design Problem (TRNDP), where route interactions can be deceptive: an extension that appears useful locally can create transfer bottlenecks, produce redundant overlap, or reduce overall throughput. To guide route construction under delayed simulator feedback, we introduce AlphaTransit, a search-based planning framework for cityscale bus network design. AlphaTransit couples Monte Carlo Tree Search (MCTS) with a neural policy-value network: the policy proposes route extensions, the value estimates downstream design quality, and search uses these predictions to refine each decision. This provides decision-time lookahead during route construction without running simulator rollouts inside the search tree. We evaluate AlphaTransit on a new Bloomington TRNDP benchmark with realistic road topology and censusderived demand, under mixed and full transit demand settings. In the Bloomington network, AlphaTransit attains the highest service rate in both demand settings, reaching 54.6% and 82.1%, respectively. Relative to reinforcement learning without search, these correspond to 9.9% and 11.4% service rate gains; relative to MCTS without learned guidance, they correspond to 2.5% and 11.2% gains. These results suggest that coupling learned guidance with MCTS is more effective than using either approach alone for transit network design. Our code and data are publicly available in https://github.com/poudel-bibek/AlphaTransit.

Ryan Wickman, Bibek Poudel, Michael Villarreal et al.

A prevalent limitation of optimizing over a single objective is that it can be misguided, becoming trapped in local optimum. This can be rectified by Quality-Diversity (QD) algorithms, where a population of high-quality and diverse solutions to a problem is preferred. Most conventional QD approaches, for example, MAP-Elites, explicitly manage a behavioral archive where solutions are broken down into predefined niches. In this work, we show that a diverse population of solutions can be found without the limitation of needing an archive or defining the range of behaviors in advance. Instead, we break down solutions into independently evolving species and use unsupervised skill discovery to learn diverse, high-performing solutions. We show that this can be done through gradient-based mutations that take on an information theoretic perspective of jointly maximizing mutual information and performance. We propose Diverse Quality Species (DQS) as an alternative to archive-based QD algorithms. We evaluate it over several simulated robotic environments and show that it can learn a diverse set of solutions from varying species. Furthermore, our results show that DQS is more sample-efficient and performant when compared to other QD algorithms. Relevant code and hyper-parameters are available at: https://github.com/rwickman/NEAT_RL.

Michael Villarreal, Bibek Poudel, Jia Pan et al.

Traffic congestion is a persistent problem in our society. Previous methods for traffic control have proven futile in alleviating current congestion levels leading researchers to explore ideas with robot vehicles given the increased emergence of vehicles with different levels of autonomy on our roads. This gives rise to mixed traffic control, where robot vehicles regulate human-driven vehicles through reinforcement learning (RL). However, most existing studies use precise observations that require domain expertise and hand engineering for each road network's observation space. Additionally, precise observations use global information, such as environment outflow, and local information, i.e., vehicle positions and velocities. Obtaining this information requires updating existing road infrastructure with vast sensor environments and communication to potentially unwilling human drivers. We consider image observations, a modality that has not been extensively explored for mixed traffic control via RL, as the alternative: 1) images do not require a complete re-imagination of the observation space from environment to environment; 2) images are ubiquitous through satellite imagery, in-car camera systems, and traffic monitoring systems; and 3) images only require communication to equipment. In this work, we show robot vehicles using image observations can achieve competitive performance to using precise information on environments, including ring, figure eight, intersection, merge, and bottleneck. In certain scenarios, our approach even outperforms using precision observations, e.g., up to 8% increase in average vehicle velocity in the merge environment, despite only using local traffic information as opposed to global traffic information.

Michael Villarreal, Bibek Poudel, Weizi Li

The surge in Reinforcement Learning (RL) applications in Intelligent Transportation Systems (ITS) has contributed to its growth as well as highlighted key challenges. However, defining objectives of RL agents in traffic control and management tasks, as well as aligning policies with these goals through an effective formulation of Markov Decision Process (MDP), can be challenging and often require domain experts in both RL and ITS. Recent advancements in Large Language Models (LLMs) such as GPT-4 highlight their broad general knowledge, reasoning capabilities, and commonsense priors across various domains. In this work, we conduct a large-scale user study involving 70 participants to investigate whether novices can leverage ChatGPT to solve complex mixed traffic control problems. Three environments are tested, including ring road, bottleneck, and intersection. We find ChatGPT has mixed results. For intersection and bottleneck, ChatGPT increases number of successful policies by 150% and 136% compared to solely beginner capabilities, with some of them even outperforming experts. However, ChatGPT does not provide consistent improvements across all scenarios.

Michael Villarreal, Bibek Poudel, Ryan Wickman et al.

With the growing use of machine learning algorithms and ubiquitous sensors, many `perception-to-control' systems are being developed and deployed. To ensure their trustworthiness, improving their robustness through adversarial training is one potential approach. We propose a gradient-free adversarial training technique, named AutoJoin, to effectively and efficiently produce robust models for image-based maneuvering. Compared to other state-of-the-art methods with testing on over 5M images, AutoJoin achieves significant performance increases up to the 40% range against perturbations while improving on clean performance up to 300%. AutoJoin is also highly efficient, saving up to 86% time per training epoch and 90% training data over other state-of-the-art techniques. The core idea of AutoJoin is to use a decoder attachment to the original regression model creating a denoising autoencoder within the architecture. This architecture allows the tasks `maneuvering' and `denoising sensor input' to be jointly learnt and reinforce each other's performance.

Bibek Poudel, Weizi Li, Kevin Heaslip

Human-driven vehicles (HVs) amplify naturally occurring perturbations in traffic, leading to congestion--a major contributor to increased fuel consumption, higher collision risks, and reduced road capacity utilization. While previous research demonstrates that Robot Vehicles (RVs) can be leveraged to mitigate these issues, most such studies rely on simulations with simplistic models of human car-following behaviors. In this work, we analyze real-world driving trajectories and extract a wide range of acceleration profiles. We then incorporates these profiles into simulations for training RVs to mitigate congestion. We evaluate the safety, efficiency, and stability of mixed traffic via comprehensive experiments conducted in two mixed traffic environments (Ring and Bottleneck) at various traffic densities, configurations, and RV penetration rates. The results show that under real-world perturbations, prior RV controllers experience performance degradation on all three objectives (sometimes even lower than 100% HVs). To address this, we introduce a reinforcement learning based RV that employs a congestion stage classifier to optimize the safety, efficiency, and stability of mixed traffic. Our RVs demonstrate significant improvements: safety by up to 66%, efficiency by up to 54%, and stability by up to 97%.

Bibek Poudel, Lei Zhu, Kevin Heaslip et al.

Modern vision systems can detect, track, and forecast urban actors at scale, yet translating perception outputs to urban design remains limited. We introduce DeCoR, a two-stage reinforcement learning framework that leverages flow observations to co-optimize crosswalk layout and network-level signal control. The design stage encodes the pedestrian network as a graph and learns a generative policy that parameterizes a Gaussian mixture model over crosswalk location and width, from which new crosswalks are sampled. For each layout, a shared control policy learns adaptive signal timings to minimize joint pedestrian and vehicle delay. On a 750 m real-world urban corridor with demand sensed from video and Wi-Fi logs, DeCoR learns a layout that reduces pedestrian arrival time to their nearest crosswalk by 23% while using fewer crosswalks than existing configurations. On the control side, DeCoR reduces pedestrian and vehicle wait time by 79% and 65%, respectively, relative to fixed-time signalization. Further, the control policy generalizes to demands outside of training and is robust to layout changes without retraining.

Adithya V. Sastry, Bibek Poudel, Weizi Li

Many robotic exploration algorithms rely on graph structures for frontier-based exploration and dynamic path planning. However, these graphs grow rapidly, accumulating redundant information and impacting performance. We present a transformer-based framework trained with Proximal Policy Optimization (PPO) to prune these graphs during exploration, limiting their growth and reducing the accumulation of excess information. The framework was evaluated on simulations of a robotic agent using Rapidly Exploring Random Trees (RRT) to carry out frontier-based exploration, where the learned policy reduces graph size by up to 96%. We find preliminary evidence that our framework learns to associate pruning decisions with exploration outcomes despite sparse, delayed reward signals. We also observe that while intelligent pruning achieves a lower rate of exploration compared to baselines, it yields the lowest standard deviation, producing the most consistent exploration across varied environments. To the best of our knowledge, these results are the first suggesting the viability of RL in sparsification of dynamic graphs used in robotic exploration algorithms.

Bibek Poudel, Abdul Basit, Muhammad Shafique

State-of-the-art 6-DoF grasp generators excel on tabletop benchmarks with overhead cameras but struggle in frontal grasping scenarios on low-cost manipulators with constrained workspaces, where kinematic limits and approach-direction constraints cause high failure rates. We address this challenge for the Unitree Z1 arm by proposing MVB-Grasp, a novel grasping stack that injects a Minimum Volume Bounding Box (MVBB) geometric prior into diffusion-based grasp generation to dramatically improve success rates in frontal, workspace-constrained settings. Our key scientific contributions are threefold: (i) an MVBB-based geometric filter that exploits oriented bounding-box face normals to reject grasps approaching through the table or misaligned with accessible object faces in O(N) time; (ii) a combined re-scoring function that blends learned discriminator scores with face-alignment geometry α=0.85, specifically calibrated for the Z1's frontal workspace and kinematic constraints; and (iii) a systematic MuJoCo evaluation protocol measuring grasp success across object types, distances, lateral positions, and pitch orientations to validate embodiment-specific performance. We implement MVB-Grasp on a Unitree Z1 arm with an Intel RealSense D405 camera, integrating YOLOv8 object detection, GraspGen for candidate generation, Principal Component Analysis (PCA)-based MVBB fitting, and inverse-kinematics trajectory planning. Experiments across 81 MuJoCo episodes (cylinder, asymmetric box, waterbottle) demonstrate that MVB-Grasp achieves 59.3% success versus 24.7% for vanilla GraspGen, a 2.4x improvement, by filtering geometrically infeasible candidates and prioritizing face-aligned grasps suited to the Z1's frontal approach constraints. Real-world trials confirm that the MVBB prior substantially improves grasp reliability on constrained, low-cost manipulators without requiring model retraining.

Bibek Poudel, Adam Cook, Sekou Traore et al.

Effective communication, specifically through documentation, is the beating heart of collaboration among contributors in software development. Recent advancements in language models (LMs) have enabled the introduction of a new type of actor in that ecosystem: LM-powered assistants capable of code generation, optimization, and maintenance. Our study investigates the efficacy of small language models (SLMs) for generating high-quality docstrings by assessing accuracy, conciseness, and clarity, benchmarking performance quantitatively through mathematical formulas and qualitatively through human evaluation using Likert scale. Further, we introduce DocuMint, as a large-scale supervised fine-tuning dataset with 100,000 samples. In quantitative experiments, Llama 3 8B achieved the best performance across all metrics, with conciseness and clarity scores of 0.605 and 64.88, respectively. However, under human evaluation, CodeGemma 7B achieved the highest overall score with an average of 8.3 out of 10 across all metrics. Fine-tuning the CodeGemma 2B model using the DocuMint dataset led to significant improvements in performance across all metrics, with gains of up to 22.5% in conciseness. The fine-tuned model and the dataset can be found in HuggingFace and the code can be found in the repository.

Bibek Poudel, Xuan Wang, Weizi Li et al.

Reinforcement learning (RL) holds significant promise for adaptive traffic signal control. While existing RL-based methods demonstrate effectiveness in reducing vehicular congestion, their predominant focus on vehicle-centric optimization leaves pedestrian mobility needs and safety challenges unaddressed. In this paper, we present a deep RL framework for adaptive control of eight traffic signals along a real-world urban corridor, jointly optimizing both pedestrian and vehicular efficiency. Our single-agent policy is trained using real-world pedestrian and vehicle demand data derived from Wi-Fi logs and video analysis. The results demonstrate significant performance improvements over traditional fixed-time signals, reducing average wait times per pedestrian and per vehicle by up to 67% and 52% respectively, while simultaneously decreasing total wait times for both groups by up to 67% and 53%. Additionally, our results demonstrate generalization capabilities across varying traffic demands, including conditions entirely unseen during training, validating RL's potential for developing transportation systems that serve all road users.

Bibek Poudel, Weizi Li

Traffic state prediction is necessary for many Intelligent Transportation Systems applications. Recent developments of the topic have focused on network-wide, multi-step prediction, where state of the art performance is achieved via deep learning models, in particular, graph neural network-based models. While the prediction accuracy of deep learning models is high, these models' robustness has raised many safety concerns, given that imperceptible perturbations added to input can substantially degrade the model performance. In this work, we propose an adversarial attack framework by treating the prediction model as a black-box, i.e., assuming no knowledge of the model architecture, training data, and (hyper)parameters. However, we assume that the adversary can oracle the prediction model with any input and obtain corresponding output. Next, the adversary can train a substitute model using input-output pairs and generate adversarial signals based on the substitute model. To test the attack effectiveness, two state of the art, graph neural network-based models (GCGRNN and DCRNN) are examined. As a result, the adversary can degrade the target model's prediction accuracy up to $54\%$. In comparison, two conventional statistical models (linear regression and historical average) are also examined. While these two models do not produce high prediction accuracy, they are either influenced negligibly (less than $3\%$) or are immune to the adversary's attack.

Bibek Poudel, Thomas Watson, Weizi Li

Autonomous micromobility has been attracting the attention of researchers and practitioners in recent years. A key component of many micro-transport vehicles is the DC motor, a complex dynamical system that is continuous and non-linear. Learning to quickly control the DC motor in the presence of disturbances and uncertainties is desired for various applications that require robustness and stability. Techniques to accomplish this task usually rely on a mathematical system model, which is often insufficient to anticipate the effects of time-varying and interrelated sources of non-linearities. While some model-free approaches have been successful at the task, they rely on massive interactions with the system and are trained in specialized hardware in order to fit a highly parameterized controller. In this work, we learn to steer a DC motor via sample-efficient reinforcement learning. Using data collected from hardware interactions in the real world, we additionally build a simulator to experiment with a wide range of parameters and learning strategies. With the best parameters found, we learn an effective control policy in one minute and 53 seconds on a simulation and in 10 minutes and 35 seconds on a physical system.