Design And Implementation Of Mobile Robot Navigation Using Spiking Neural Network On A Neuromorphic Processor

Wibowo, Bambang (2024) Design And Implementation Of Mobile Robot Navigation Using Spiking Neural Network On A Neuromorphic Processor. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5022201163-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (2MB) | Request a copy

Abstract

Spiking neural networks (SNN) adalah artificial neural networks (ANN) yang paling mendekati secara biologis dengan otak sebenarnya. SNN juga memiliki banyak keuntungan daripada ANN biasa, seperti hemat daya, dan pengolahan informasi dengan cepat dan efisien. SNN sudah diterapkan pada navigasi mobile robot menggunakan aturan pembelajaran seperti Spike-Timing-Dependent Plasticity (STDP), dan menggunakan model neuron seperti Leaky-Integrate-and-Fire (LIF). Dalam Tugas Akhir ini telah dikembangkan mobile robot navigation menggunakan neuromorphic processor berbasis field programmable gate array (FPGA) dan SNN. Robot ini dilengkapi dengan sensor lidar yang kompatibel dengan arsitektur SNN untuk melakukan navigasi menuju lokasi target dan menghindari rintangan secara langsung. Pengujian di simulasi dan dunia nyata menunjukkan tingkat keberhasilan yang tinggi. Goal Approaching (GA) SNN mencapai tingkat keberhasilan 95% dalam mencapai target, sedangkan Obstacle Avoidance (OA) SNN berhasil mencapai tingkat keberhasilan 80% dalam menghindari rintangan. Secara keseluruhan, mereka mencapai tingkat keberhasilan 90%. Implementasi pada CPU dan FPGA sama-sama efektif, namun FPGA menghadapi masalah komunikasi yang meningkatkan inference time. Eksperimen konsumsi daya menunjukkan perbedaan minimal antara implementasi CPU dan CPU+FPGA, dengan kondisi CPU saja memiliki konsumsi daya rata-rata yang sedikit lebih tinggi.
============================================================
Spiking neural networks (SNN) are artificial neural networks (ANN) that closely mimic the real brain. SNNs have some advantages such as energy efficiency and computational speed. SNN has been applied to mobile robot navigation using Spike-Timing-Dependent Plasticity (STDP) as the learning rule and using neuron model such as Leaky-Integrate-and-Fire (LIF). This project has developed a mobile robot that navigates using a neuromorphic processor based on a field programmable gate array (FPGA) and SNN. The robot is equipped with sensors compatible with the SNN architecture to navigate to target locations and avoid obstacles in real time. Extensive testing in simulated and real-world environments demonstrated high success rates. The Goal Approaching (GA) SNN achieved a 95% success rate in reaching targets, while the Obstacle Avoidance (OA) SNN managed an 80% success rate in avoiding obstacles. Combined, they achieved a 90% success rate. Both CPU and FPGA implementations were effective, but the FPGA faced communication overhead issues that increased inference times. Power consumption experiments showed minimal differences between the CPU and CPU+FPGA implementations, with the CPU-only condition having a slightly higher average power consumption. Future work will focus on optimizing communication protocols between the CPU and FPGA to reduce inference times, refining the robot's turning mechanism, and extensive testing in dynamic environments to improve performance.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Kata kunci: SNN, mobile robot, neuromorphic processor, FPGA, LIF, STDP. ============================================================ Keywords: SNN, mobile robot, neuromorphic processor, FPGA, LIF, STDP.
Subjects:	T Technology > TA Engineering (General). Civil engineering (General) > TA167.5 Neurotechnology. Neuroadaptive systems
Depositing User:	Bambang Triartho Wibowo
Date Deposited:	05 Aug 2024 01:34
Last Modified:	05 Aug 2024 01:34
URI:	http://repository.its.ac.id/id/eprint/111188

Actions (login required)

View Item