Learning-based control
Safe navigation and continuous control for mobile robots, trained end-to-end with deep reinforcement learning.
PPODDPGSACROSGazebo
Open to PhD positions & research collaborations
Robotics · Perception · Deep RL
Hamid Taheri
I build learning-based perception and control systems
for intelligent robots and real-world AI.
Working at the intersection of deep reinforcement learning, computer vision, robotics, and production AI — from safe mobile-robot navigation to real-time perception infrastructure.
sensor: LiDAR · 36 beams
policy: active · obstacle-aware
pos [0.00, 0.00] · v 0.00
policy: active · obstacle-aware
pos [0.00, 0.00] · v 0.00
scroll
11open-source projects
2arXiv preprints
★53navbot_ppo stars
Top 0.2%national M.Sc. rank
Research identity
Three threads, one system.
My work spans the full loop of an intelligent machine — sensing the world, learning a policy, and shipping it reliably. These threads reinforce each other.
Perception & 3D vision
Object & anomaly detection, OCR, lane and 3D scene understanding — from point clouds to industrial imaging.
PointNet++DGCNNViTYOLOOpenCV
Reliable AI systems
Production-grade ML: GPU inference, backend services, containers and CI/CD that move research into the real world.
TensorRTDockerKubernetesC/C++Linux
Featured research
Published work.
Peer-reviewable contributions in robot navigation and efficient perception. Full text on arXiv.
Deep Reinforcement Learning with Enhanced PPO for Safe Mobile Robot Navigation
This study trains a deep reinforcement learning agent to enable a wheeled mobile robot to navigate autonomously and collision-free toward a target in complex, obstacle-rich environments. Using only raw LiDAR observations and the goal pose, a deep neural network maps perception directly to continuous velocity commands. The work compares DDPG and PPO and introduces an enhanced PPO network architecture together with a tailored reward function, validated in both obstacle and obstacle-free Gazebo environments.
Hamid Taheri, Seyed Rasoul Hosseini, Mohammad Ali Nekoui
ENet-21: An Optimized Light CNN Structure for Lane Detection
This study develops an optimized, lightweight CNN for lane detection that combines binary segmentation with Affinity Fields. Lane instances are recovered by clustering the segmentation and affinity outputs, which lets the method handle a varying number of lanes and lane-change scenarios without a fixed lane-count assumption. Using a less complex network than existing approaches, the method is demonstrated on the TuSimple dataset.
Seyed Rasoul Hosseini, Hamid Taheri, Mohammad Teshnehlab
Selected work
Evidence in code.
Research, perception, and systems work — built, trained, and shipped.
Flagship · Deep RL
Featured work
navbot_ppo
Mapless mobile-robot navigation: deep-RL motion planners — PPO, DDPG, SAC and TD3 — that drive a TurtleBot to a goal from sparse LiDAR and the target pose, with no SLAM and no map.
★ 53GitHub stars
16-DLiDAR state
maplessno SLAM
Point clouds
7
3D Scene Understanding
From-scratch reimplementations of PointNet, PointNet++, DGCNN and PointMLP for point-cloud classification and part segmentation.
3D_Bbox_detection
RGB-D
1
3D Object Detection on RGB-D
An end-to-end 3D object-detection pipeline on RGB-D data, comparing a from-scratch model against a fine-tuned EfficientNet-B3 backbone under a small-data regime.
ViT-AnomalyDetection
Inspection
5
ViT-AnomalyDetection
A Vision-Transformer autoencoder for industrial inspection: reconstruct the image, then localize defects from patch-wise reconstruction error.
YoloV3_from_scratch
COCO · 90% mAP
YOLOv3 Object Detection
A real-time object-detection pipeline built with YOLOv3 and non-max suppression on the COCO dataset.
Ray tracer · C
1
MiniRT — Ray Tracer in C
A from-scratch ray tracer in C (miniLibX): ray–object intersections, Phong shading, and an .rt scene parser producing photorealistic renders.
Recent
News & updates.
Mar 2026
Started as Applied AI Software Engineer at amitego Engineering, building VMask — real-time OCR and computer-vision privacy masking for remote desktop sessions.
Nov 2025
Joined PreZero (Schwarz Group) as an AI/ML & Full-Stack Innovation Intern, working on logistics route optimization and predictive container-pickup planning.
Jun 2025
Top Performer at the Audi XL2 DeepRacer reinforcement-learning competition — virtual and on-site autonomous-navigation racing.
Aug 2024
Revised preprints posted: Enhanced-PPO safe navigation (arXiv:2405.16266) and ENet-21 lane detection (arXiv:2403.19782), both under journal review.