Autonomous Solutions

Technological advances in the automotive industry have increased the complexity of vehicle testing and simultaneously require sophisticated methods to ensure safety, accuracy and reproducibility. Testing on proving grounds is essential to validate advanced systems at performance limits or for durability testing under controlled and repeatable conditions. However, these tests require precise planning, compliance and accuracy, which are error-prone and physically demanding for vehicles and drivers. Automation helps to overcome these challenges. This work presents an innovative approach using an Intelligent Test Coordinator (ITC) and Robot Driving System (RDS) for test vehicles to automate test preparation and execution. The ITC uses a Large Language Model and error feedback to automate the design of test procedures and generate precise RDS instructions for real-world test scenarios. Simulations validate test setups before execution, and vehicles equipped with RDS perform driverless testing. The performance of the ITC is validated by creating 240 test setups for different driving scenarios. The robot-guided vehicles perform real-world tests based on test setups on various tracks, including flat asphalt surfaces and handling tracks. These tests demonstrate that the proposed approach outperforms human-driven vehicles in terms of safety, accuracy and repeatability. The proposed approach to automation enhances testing efficiency and facilitates comprehensive validation of vehicle safety and performance.

Previous studies established strong links between morphological characteristics of mammalian hindlimb muscles and their sensorimotor functions during locomotion. Less is known about the role of forelimb morphology in motor outputs and generation of sensory signals. Here, we measured morphological characteristics of 46 forelimb muscles from 6 cats. These characteristics included muscle attachments, physiological cross-sectional area (PCSA), fascicle length, etc. We also recorded full-body mechanics and EMG activity of forelimb muscles during level overground and treadmill locomotion in 7 and 16 adult cats of either sex, respectively. We computed forelimb muscle forces along with force- and length-dependent sensory signals mapped onto corresponding cervical spinal segments. We found that patterns of computed muscle forces and afferent activities were strongly affected by the muscles moment arm, PCSA, and fascicle length. Morphology of the shoulder muscles suggests distinct roles of the forelimbs in lateral force production and movements. Patterns of length-dependent sensory activity of muscles with long fibers (brachioradialis, extensor carpi radialis) closely matched patterns of overall forelimb length, whereas the activity pattern of biceps brachii matched forelimb orientation. We conclude that cat forelimb muscle morphology contributes substantially to locomotor function, particularly to control lateral stability and turning, rather than propulsion.

In this paper, we consider the computation of robust N-step backward reachable sets for state- and input con- strained linear time-varying systems with additive uncertainty. We propose a method to compute a linear, time-varying control law that maximizes the volume of the robust N-step reachable set for the closed-loop system. The proposed method is an extension of recent developments and involves the recursive solution of N semi-definite programs (SDP). We demonstrate the performance of the proposed method on the lateral control problem for emergency maneuvers of autonomous vehicles and compare it to results obtained when backward reachability is applied to the same system and a naively designed controller.

The contemporary processors are unable to meet the increasing data-intensive and computation-demanding requirements in autonomous vehicle software applications. Recently, the new Heterogeneous System Architecture (HSA) has emerged as a promising solution to meet these requirements. The HSA reduces the latency of data exchange between the compute units and cache-coherent shared memory, which is not supported by the non-HSA compliant heterogeneous platforms with acceleration support. The main goal of the paper is to investigate the performance gain by the HSA and conduct a comparative evaluation of the HSA and non-HSA compliant heterogeneous platforms. The paper aims at evaluating these platforms by using two computation-intensive software functions in autonomous vehicles, namely the object detection and vehicle movement. In order to achieve this goal, the CUDA-accelerated source code of the functions is ported from a non-HSA compliant heterogeneous platform to the HSA platform. In this regard, the paper presents the architecture of a proof-of-concept prototype and provides evaluation using the prototype.

Over the period 18–21 March, 2017, 1-hour waveforms in the 60–120 Hz band were transmitted from a HLF-6A source deployed at 300 m depth in the deep northeast Pacific Ocean. These transmissions were recorded by two vertical hydrophone line arrays at various ranges in the deep ocean, GPS-equipped sonobuoys deployed on the continental shelf, and a bottom-mounted hydrophone in 900-m water at the western edge of Monterey Bay operated by the Monterey Bay Aquarium and Research Institute (MBARI). Although the source-receiver ranges to the sonobuoys and the MBARI hydrophone were approximately the same, the bathymetry profile up the continental slope to the sonobuoy location was significantly steeper than to the MBARI hydrophone. Only the upper part of the frequency band, above 90–95 Hz, is received with good signal-to-noise ratio, illustrating the high-pass temporal filtering of upslope propagation. Upslope propagation also acts as a low-pass spatial filter, allowing only lower-order modes to propagate onto the shelf. Numerical modeling is used to examine the predictability of the measured travel times and multipath arrival structure.