Searcharxiv⌕ Search

arXiv subjects

John Walsh

Publications and source records attributed to John Walsh.

10 recordsLinked to original sources
arXiv2026

Fuzzy Encoding-Decoding to Improve Spiking Q-Learning Performance in Autonomous Driving

This paper develops an end-to-end fuzzy encoder-decoder architecture for enhancing vision-based multi-modal deep spiking Q-networks in autonomous driving. The method addresses two core limitations of spiking reinforcement learning: information loss stemming from the conversion of dense visual inputs into sparse spike trains, and the limited representational capacity of spike-based value functions, which often yields weakly discriminative Q-value estimates. The encoder introduces trainable fuzzy membership functions to generate expressive, population-based spike representations, and the decoder uses a lightweight neural decoder to reconstruct continuous Q-values from spiking outputs. Experiments on the HighwayEnv benchmark show that the proposed architecture substantially improves decision-making accuracy and closes the performance gap between spiking and non-spiking multi-modal Q-networks. The results highlight the potential of this framework for efficient and real-time autonomous driving with spiking neural networks.

cs.NE↗
arXiv2025

New Spiking Architecture for Multi-Modal Decision-Making in Autonomous Vehicles

This work proposes an end-to-end multi-modal reinforcement learning framework for high-level decision-making in autonomous vehicles. The framework integrates heterogeneous sensory input, including camera images, LiDAR point clouds, and vehicle heading information, through a cross-attention transformer-based perception module. Although transformers have become the backbone of modern multi-modal architectures, their high computational cost limits their deployment in resource-constrained edge environments. To overcome this challenge, we propose a spiking temporal-aware transformer-like architecture that uses ternary spiking neurons for computationally efficient multi-modal fusion. Comprehensive evaluations across multiple tasks in the Highway Environment demonstrate the effectiveness and efficiency of the proposed approach for real-time autonomous decision-making.

cs.LG↗
arXiv2025

Improving Performance of Spike-based Deep Q-Learning using Ternary Neurons

We propose a new ternary spiking neuron model to improve the representation capacity of binary spiking neurons in deep Q-learning. Although a ternary neuron model has recently been introduced to overcome the limited representation capacity offered by binary spiking neurons, we show that its performance is worse than that of binary models in deep Q-learning tasks, contradicting previous findings from recent studies. Through mathematical and empirical analysis, we hypothesize that gradient estimation bias during training is the underlying cause. The proposed ternary spiking neuron model mitigates this issue by reducing the estimation bias. We use the proposed ternary spiking neuron as the fundamental computing unit in a deep spiking Q-learning network, which we call the deep asymmetric ternary spiking Q-network (DATSQN), and evaluate the network's performance in seven Atari games from the Gym environment. The results show that the proposed ternary spiking neuron mitigates the performance degradation of ternary neurons in DQN tasks and improves the mean game score relative to the binary baseline under the evaluation settings used in this paper.

cs.LG↗
arXiv2020

Estimation and Decomposition of Rack Force for Driving on Uneven Roads

The force transmitted from the front tires to the steering rack of a vehicle, called the rack force, plays an important role in the function of electric power steering (EPS) systems. Estimates of rack force can be used by EPS to attenuate road feedback and reduce driver effort. Further, estimates of the components of rack force (arising, for example, due to steering angle and road profile) can be used to separately compensate for each component and thereby enhance steering feel. In this paper, we present three vehicle and tire model-based rack force estimators that utilize sensed steering angle and road profile to estimate total rack force and individual components of rack force. We test and compare the real-time performance of the estimators by performing driving experiments with non-aggressive and aggressive steering maneuvers on roads with low and high frequency profile variations. The results indicate that for aggressive maneuvers the estimators using non-linear tire models produce more accurate rack force estimates. Moreover, only the estimator that incorporates a semi-empirical Rigid Ring tire model is able to capture rack force variation for driving on a road with high frequency profile variation. Finally, we present results from a simulation study to validate the component-wise estimates of rack force.

eess.SY↗
arXiv2020

Rack Force Estimation for Driving on Uneven Road Surfaces

The force transmitted from the front tires and tie rods to the steering rack of a vehicle, called the rack force, significantly influences the torque experienced by a driver at the steering wheel. As a result, estimates of rack force are used in a wide variety of advanced driver assist systems. Existing methods for producing rack force estimates are either susceptible to steering system disturbances or are only applicable for driving on roads with low frequency profile variations such as road slopes. In this paper we present a model that can produce disturbance-free rack force estimates for driving on roads with high frequency profile variations, such as road cleats and potholes, in addition to roads with low frequency profile variations. We validate the estimation accuracy of our model by presenting results from two driving experiments that were performed on test tracks with known low and high frequency road profile variations. We further demonstrate the merits of our model relative to the existing models by comparing the various estimates to rack force measurements obtained using a sensor mounted in the test vehicle.

eess.SY↗
arXiv2020

The Locus Algorithm III: A Grid Computing system to generate catalogues of optimised pointings for Differential Photometry

This paper discusses the hardware and software components of the Grid Computing system used to implement the Locus Algorithm to identify optimum pointings for differential photometry of 61,662,376 stars and 23,799 quasars. The scale of the data, together with initial operational assessments demanded a High Performance Computing (HPC) system to complete the data analysis. Grid computing was chosen as the HPC solution as the optimum choice available within this project. The physical and logical structure of the National Grid computing Infrastructure informed the approach that was taken. That approach was one of layered separation of the different project components to enable maximum flexibility and extensibility.

astro-ph.IM↗
arXiv2020

The Locus Algorithm IV: Performance metrics of a grid computing system used to create catalogues of optimised pointings

This paper discusses the requirements for and performance metrics of the the Grid Computing system used to implement the Locus Algorithm to identify optimum pointings for differential photometry of 61,662,376 stars and 23,779 quasars. Initial operational tests indicated a need for a software system to analyse the data and a High Performance Computing system to run that software in a scalable manner. Practical assessments of the performance of the software in a serial computing environment were used to provide a benchmark against which the performance metrics of the HPC solution could be compared, as well as to indicate any bottlenecks in performance. These performance metrics indicated a distinct split in the performance dictated more by differences in the input data than by differences in the design of the systems used. This indicates a need for experimental analysis of system performance, and suggests that algorithmic complexity analyses may lead to incorrect or naive conclusions, especially in systems with high data I/O overhead such as grid computing. Further, it implies that systems which reduce or eliminate this bottleneck such as in-memory processing could lead to a substantial increase in performance.

astro-ph.IM↗
arXiv2014

Inclusive $B\to X_sγ$ and $B\to X_s\ell^+\ell^-$ at the B factories

I report here recent measurements of observables from the inclusive decays $B\to X_sγ$ and $B\to X_s\ell^+\ell^-$. Included are measurements of the branching fractions and CP asymmetries for both channels, as well as the forward-backward lepton asymmetry in inclusive $B\to X_s\ell^+\ell^-$ decays, which is the first measurement of this quantity.

hep-ex↗
arXiv2005

Semileptonic and Electroweak Penguin Results from BABAR

We report recent results from the BABAR experiment on semileptonic charmless B-meson decays and electroweak penguin processes. Semileptonic charmless decays are used to determine |Vub| and the exclusive modes considered here also begin to constrain QCD-lattice form factor calculations. Radiative penguin decays are both sensitive to physics beyond the Standard Model and can be used to extract Heavy Quark parameters related to the b-quark mass and its motion inside the hadron.

hep-ex↗