P3Net: PointNet-based Path Planning on FPGA

Sugiura, Keisuke; Matsutani, Hiroki

doi:10.1109/icfpt56656.2022.9974251

Cited by 4 publications

(5 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [36], an RTL design of a Graph Neural Networkbased path explorer rapidly evaluates priority scores for edges in a random geometric graph, and edges with high priority are selected to form a path. This paper extends our previous work [34]; instead of only accelerating the DNN inference in MPNet, we implement the whole bidirectional planning algorithm on FPGA. In addition, we derive a new path planning method, P3Net, to achieve both higher success rate and speedup.…”

Section: Hardware Acceleration Of Path Planningmentioning

confidence: 58%

“…Only a few works [34], [35], [36] have considered the hardware acceleration of neural planners. Huang et al [35] presents an accelerator for a sampling-based method with a CNN model, which produces a probability map given an image of the environment for sampling the next robot position.…”

Section: Hardware Acceleration Of Path Planningmentioning

confidence: 99%

“…1-2, the bidirectional neural planning (NeuralPlannerEx) is at the core of P3Net and thus has a significant impact on the overall performance. In contrast to the previous work [34], which only implements PNet inference on the custom IP core, NeuralPlanner covers the entire NeuralPlannerEx algorithm (Alg. 3).…”

Section: Neuralplanner Modulementioning

confidence: 99%

“…5.2.1) is set sufficiently small to prevent false negatives. Following the previous work [34], we create another IP core for P3Net, P3NetCore-NN, that only implements the {E, P}NetLite inference.…”

Section: Implementation Detailsmentioning

confidence: 99%

See 3 more Smart Citations

An Integrated FPGA Accelerator for Deep Learning-Based 2D/3D Path Planning

Sugiura,

Matsutani

2024

IEEE Trans. Comput.

Self Cite

View full text Add to dashboard Cite

Path planning is a crucial component for realizing the autonomy of mobile robots. However, due to limited computational resources on mobile robots, it remains challenging to deploy state-of-the-art methods and achieve real-time performance. To address this, we propose P3Net (PointNet-based Path Planning Networks), a lightweight deep-learning-based method for 2D/3D path planning, and design an IP core (P3NetCore) targeting FPGA SoCs (Xilinx ZCU104). P3Net improves the algorithm and model architecture of the recently-proposed MPNet. P3Net employs an encoder with a PointNet backbone and a lightweight planning network in order to extract robust point cloud features and sample path points from a promising region. P3NetCore is comprised of the fully-pipelined point cloud encoder, batched bidirectional path planner, and parallel collision checker, to cover most part of the algorithm. On the 2D (3D) datasets, P3Net with the IP core runs 30. 52-186.36x and 7.68-143.62x (15.69-93.26x and 5.30-45.27x) faster than ARM Cortex CPU and Nvidia Jetson while only consuming 0.255W (0.809W), and is up to 1278.14x (455.34x) power-efficient than the workstation. P3Net improves the success rate by up to 28.2% and plans a near-optimal path, leading to a significantly better tradeoff between computation and solution quality than MPNet and the state-of-the-art sampling-based methods.

show abstract

Section: Hardware Acceleration Of Path Planningmentioning

confidence: 58%

Section: Hardware Acceleration Of Path Planningmentioning

confidence: 99%

Section: Neuralplanner Modulementioning

confidence: 99%

Section: Implementation Detailsmentioning

confidence: 99%

See 2 more Smart Citations

An Integrated FPGA Accelerator for Deep Learning-Based 2D/3D Path Planning

Sugiura,

Matsutani

2024

IEEE Trans. Comput.

Self Cite

View full text Add to dashboard Cite

show abstract

“…The work [38] presented a hardware implementation of the PointNet architecture, achieving a processing time of 19.8 ms for 4096 points from LiDAR on the AMD Xilinx's ZCU104 platform. Similarly, [39] implemented the PointNet architecture on an FPGA for pathfinding and obstacle avoidance in a cloud of 1400 points. However, both works processed data without creating a graph; neither vertices were interconnected by edges nor their relationships were defined.…”

Section: Workmentioning

confidence: 99%

Traffic sign detection and recognition using event camera image reconstruction

Jeziorek¹,

Kryjak²

2022

Preprint

View full text Add to dashboard Cite

<p>This paper presents a method for detection and recognition of traffic signs based on information extracted from an event camera. The solution used a FireNet deep convolutional neural network to reconstruct events into greyscale frames. Two YOLOv4 network models were trained, one based on greyscale images and the other on colour images. The best result was achieved for the model trained on the basis of greyscale images, achieving an efficiency of 87.03%.</p>

show abstract