An Thai Le

An Thai Le

Auto-generated profile, based on publicly available information

An Thai Le

VinUniversity

Research Focus & Metrics

Academic Career Length

11Years (2016-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
28 July 2025 in IEEE Transactions on Robotics
Verified
Journal Article
Closed access
JCJoão Carvalho
João Carvalho
ALAn Thai Le
An Thai Le
PKPiotr Kicki
Piotr Kicki
DKDorothea Koert
Dorothea Koert
Jan PetersJan Peters
Jan Peters
The performance of optimization-based robot motion planning algorithms is highly dependent on the initial solutions, commonly obtained by running a sampling-based planner to obtain a collision-free path. However, these methods can be slow in high-dimensional and complex scenes and produce nonsmooth solutions. Given previously solved path-planning problems, it is highly desirable to learn their distribution and use it as a prior for new similar problems. Several works propose utilizing this prior to bootstrap the motion planning problem, either by sampling initial solutions from it, or using its distribution in a maximum-a-posterior formulation for trajectory optimization. In this work, we introduce motion planning diffusion (MPD), an algorithm that learns trajectory distribution priors with diffusion models. These generative models have shown increasing success in encoding multimodal data and have desirable properties for gradient-based motion planning, such as cost guidance. Given a motion planning problem, we construct a cost function and sample from the posterior distribution using the learned prior combined with the cost function gradients during the denoising process. Instead of learning the prior on all trajectory waypoints, we propose learning a lower dimensional representation of a trajectory using linear motion primitives, particularly B-spline curves. This parametrization guarantees that the generated trajectory is smooth, can be interpolated at higher frequencies, and needs fewer parameters than a dense waypoint representation. We demonstrate the results of our method ranging from simple 2-D to more complex tasks using a 7-DOF robot arm manipulator. In addition to learning from simulated data, we also use human demonstrations on a real-world pick-and-place task. The experiment results show that diffusion models are strong priors for encoding multimodal trajectory distributions for optimization-based motion planning.

26 September 2018
Verified
Book Chapter
Open access
PDF file
ALAn T. Le
An T. Le
TLThan Le
Than Le
In this chapter, we present one of the most crucial branches in motion planning: search-based planning and replanning algorithms. This research branch involves two key points: first, representing traverse environment information as discrete graph form, in particular, occupancy grid cost map at arbitrary resolution, and, second, path planning algorithms calculate paths on these graphs from start to goal by propagating cost associated with each vertex in graph. The chapter will guide researcher through the foundation of motion planning concept, the history of search-based path planning and then focus on the evolution of state-of-the-art incremental, heuristic, anytime algorithm families that are currently applied on practical robot rover. The comparison experiment between algorithm families is demonstrated in terms of performance and optimality. The future of search-based path planning and motion planning in general is also discussed.

30 May 2025 in IEEE Robotics and Automation Letters
Verified
Journal Article
Closed access
ALAn T. Le
An T. Le
KPKay Pompetzki
Kay Pompetzki
JCJoão Carvalho
João Carvalho
JWJoe Watson
Joe Watson
JUJulen Urain
Julen Urain
KHKay Hansel
Kay Hansel
Batch planning is increasingly necessary to quickly produce diverse and quality motion plans for downstream learning applications, such as distillation and imitation learning. This paper presents Global Tensor Motion Planning (GTMP)—a sampling-based motion planning algorithm comprising only tensor operations. We introduce a novel discretization structure represented as a random multipartite graph, enabling efficient vectorized sampling, collision checking, and search. We provide a theoretical investigation showing that GTMP exhibits probabilistic completeness while supporting modern GPU/TPU. Additionally, by incorporating smooth structures into the multipartite graph, GTMP directly plans smooth splines without requiring gradient-based optimization. Experiments on lidar-scanned occupancy maps and the MotionBenchMarker dataset demonstrate GTMP's computation efficiency in batch planning compared to baselines, underscoring GTMP's potential as a robust, scalable planner for diverse applications and large-scale robot learning tasks.

1 April 2017
Verified
Conference Paper
Closed access
ALAn T. Le
An T. Le
MBMinh Q. Bui
Minh Q. Bui
TLThan Le
Than Le
NPNauth Peter
Nauth Peter
One of the most vital aspects of navigation problem is path planning, in which we must be able to find the optimal path with minimum total transitions cost from the start to goal. This paper introduces the state-of-the-art D*Lite algorithm and presents the efficiency problem of D* Lite in complex environments, particularly in dynamic maze-like environments, where the effect of changes is costly. We propose the new algorithm D* Lite with Reset (D*LR) with two different criteria for resetting. We prove experimentally the performance improvement of D*LR over D* Lite in complex environments. Further implementation of D*LR on both Processing simulator and robotics framework ROS is also presented.

1 July 2017
Verified
Conference Paper
Closed access
TLThan Le
Than Le
ALAn T. Le
An T. Le
DNDuy T. Nguyen
Duy T. Nguyen
This paper is proposal with regarding reinforcement learning and robotics. It contains our study so far about reinforcement learning problem and Q-learning - one of the methods to solve it. The method is tested both by running on a simulation and on NAO robot. Both are written in high programming language. Since the testing are also done on NAO robot. This paper also includes our understanding about NAO robot and Robotic Operating System (ROS), and our approach to apply Q-learning on NAO robot. In the end, we have been successfully tested Q-learning method and apply it to NAO robot in real-world environment.

23 August 2016
Verified
Book Chapter
Closed access
PDF file
KDKhiem N. Doan
Khiem N. Doan
ALAn T. Le
An T. Le
TLThan Le
Than Le
NPNauth Peter
Nauth Peter