Search Results

The SAE AutoDrive Challenge II is a four-year collegiate competition dedicated to developing a Level 4 autonomous vehicle by 2025. In January 2023, the participating teams each received a Chevy Bolt EUV. Within a span of five months, the second phase of the competition took place in Ann Arbor, MI. The authors of this contribution, who participated in this event as team Wisconsin Autonomous representing the University of Wisconsin–Madison, secured second place in static events and third place in dynamic events. This has been accomplished by reducing reliance on the actual vehicle platform and instead leveraging physical analogs and simulation. This paper outlines the software and hardware infrastructure of the competing vehicle, touching on issues pertaining sensors, hardware, and the software architecture employed on the autonomous vehicle. We discuss the LiDAR-camera fusion approach for object detection and the three-tier route planning and following systems.

Lithium-ion and Lithium polymer batteries are fast becoming ubiquitous in high-discharge rate applications for military and non-military systems. Applications such as small aerial vehicles and energy transfer systems can often function at C-rates greater than 1. To maximize system endurance and battery health, there is a need for models capable of precisely estimating the battery state-of-charge (SoC) under all temperature and loading conditions. However, the ability to perform state estimation consistently and accurately to within 1% error has remained unsolved. Doing so can offer enhanced endurance, safety, reliability, and planning, and additionally, simplify energy management. Therefore, the work presented in this paper aims to study and develop experimentally validated mathematical models capable of high-accuracy battery SoC estimation.

Identifying edge cases for designed algorithms is critical for functional safety in autonomous driving deployment. In order to find the feasible boundary of designed algorithms, simulations are heavily used. However, simulations for autonomous driving validation are expensive due to the requirement of visual rendering, physical simulation, and AI agents. In this case, common sampling techniques, such as Monte Carlo Sampling, become computationally expensive due to their sample inefficiency. To improve sample efficiency and minimize the number of simulations, we propose a tailored active learning approach combining the Support Vector Machine (SVM) and the Gaussian Process Regressor (GPR). The SVM learns the feasible boundary iteratively with a new sampling point via active learning. Active Learning is achieved by using the information of the decision boundary of the current SVM and the uncertainty metric calculated by the GPR.

Traffic state identification is a key problem in intelligent transportation system. As a new technology, connected and automated vehicle can play a role of identifying traffic state with the installation of onboard sensors. However, research of lane level traffic state identification is relatively lacked. Identifying lane level traffic state is helpful to lane selection in the process of driving and trajectory planning. In addition, traffic state identification precision with low penetration of connected and automated vehicles is relatively low. To fill this gap, this paper proposes a novel method of identifying traffic state in the presence of connected and automated vehicles with low penetration rate. Assuming connected and automated vehicles can obtain information of surrounding vehicles’, we use the perceptible information to estimate imperceptible information, then traffic state of road section can be inferred.

Appropriate spray modeling in multidimensional simulations of diesel engines is well known to affect the overall accuracy of the results. More and more accurate models are being developed to deal with drop dynamics, breakup, collisions, and vaporization/multiphase processes; the latter ones being the most computationally demanding. In fact, in parallel calculations, the droplets occupy a physical region of the in-cylinder domain, which is generally very different than the topology-driven finite-volume mesh decomposition. This makes the CPU decomposition of the spray cloud severely uneven when many CPUs are employed, yielding poor parallel performance of the spray computation. Furthermore, mesh-independent models such as collision calculations require checking of each possible droplet pair, which leads to a practically intractable O(np2/2) computational cost, np being the total number of droplets in the spray cloud, and additional overhead for parallel communications.

We present an approach in which an open-source software infrastructure is used for testing the behavior of autonomous vehicles through computer simulation. This software infrastructure is called CAVE, from Connected Autonomous Vehicle Emulator. As a software platform that allows rapid, low-cost and risk-free testing of novel designs, methods and software components, CAVE accelerates and democratizes research and development activities in the field of autonomous navigation.

Past research has shown that reactivity controlled compression ignition (RCCI) combustion offers efficiency and NOx and soot advantages over conventional diesel combustion at mid load conditions. However, at high load and low speed conditions, the chemistry timescale of the fuel shortens and the engine timescale lengthens. This mismatch in timescales makes operation at high load and low speed conditions difficult. High levels of exhaust gas recirculation (EGR) can be used to extend the chemistry timescales; however, this comes at the penalty of increased pumping losses. In the present study, targeting the high load - low speed regime, computational optimizations of RCCI combustion were performed at 20 bar gross indicated mean effective pressure (IMEP) and 1300 rev/min. The two fuels used for the study were gasoline (low reactivity) and diesel (high reactivity).

An investigation of high speed direct injection (DI) compression ignition (CI) engine combustion fueled with gasoline injected using a triple-pulse strategy in the low temperature combustion (LTC) regime is presented. This work aims to extend the operation ranges for a light-duty diesel engine, operating on gasoline, that have been identified in previous work via extended controllability of the injection process. The single-cylinder engine (SCE) was operated at full load (16 bar IMEP, 2500 rev/min) and computational simulations of the in-cylinder processes were performed using a multi-dimensional CFD code, KIVA-ERC-Chemkin, that features improved sub-models and the Chemkin library. The oxidation chemistry of the fuel was calculated using a reduced mechanism for primary reference fuel combustion chosen to match ignition characteristics of the gasoline fuel used for the SCE experiments.

Multi-zone CFD simulations with detailed kinetics were used to model iso-octane HCCI experiments performed on a single-cylinder research engine. The modeling goals were to validate the method (multi-zone combustion modeling) and the reaction mechanism (LLNL 857 species iso-octane) by comparing model results to detailed exhaust speciation data, which was obtained with gas chromatography. The model is compared to experiments run at 1200 RPM and 1.35 bar boost pressure over an equivalence ratio range from 0.08 to 0.28. Fuel was introduced far upstream to ensure fuel and air homogeneity prior to entering the 13.8:1 compression ratio, shallow-bowl combustion chamber of this 4-stroke engine. The CFD grid incorporated a very detailed representation of the crevices, including the top-land ring crevice and head-gasket crevice. The ring crevice is resolved all the way into the ring pocket volume. The detailed grid was required to capture regions where emission species are formed and retained.

Optimal fuel injection strategies are obtained with a micro-genetic algorithm and an adaptive gradient method for a nonroad, medium-speed DI diesel engine equipped with a multi-orifice, asynchronous fuel injection system. The gradient optimization utilizes a fast-converging backtracking algorithm and an adaptive cost function which is based on the penalty method, where the penalty coefficient is increased after every line search. The micro-genetic algorithm uses parameter combinations of the best two individuals in each generation until a local convergence is achieved, and then generates a random population to continue the global search. The optimizations have been performed for a two pulse fuel injection strategy where the optimization parameters are the injection timings and the nozzle orifice diameters.

A previously developed CFD-based optimization tool is utilized to find optimal engine operating conditions with respect to fuel consumption and emissions. The optimization algorithm employed is based on the steepest descent method where an adaptive cost function is minimized along each line search using an effective backtracking strategy. The adaptive cost function is based on the penalty method, where the penalty coefficient is increased after every line search. The parameter space is normalized and, thus, the optimization occurs over the unit cube in higher-dimensional space. The application of this optimization tool is demonstrated for the Sulzer S20, a central-injection, non-road DI diesel engine. The optimization parameters are the start of injection of the two pulses of a split injection system, the duration of each pulse, the exhaust gas recirculation rate, the boost pressure and the compression ratio.

A global optimization method has been developed for an engine simulation code and utilized in the search of optimal fuel injection strategies. This method uses a Lagrange interpolation function which interpolates engine output data generated at the vertices and the intermediate points of the input parameters. This interpolation function is then used to find a global minimum over the entire parameter set, which in turn becomes the starting point of a CFD-based optimization. The CFD optimization is based on a steepest descent method with an adaptive cost function, where the line searches are performed with a fast-converging backtracking algorithm. The adaptive cost function is based on the penalty method, where the penalty coefficient is increased after every line search. The parameter space is normalized and, thus, the optimization occurs over the unit cube in higher-dimensional space.

Engine-out CO emission and fuel conversion efficiency were measured in a highly-dilute, low-temperature diesel combustion regime over a swirl ratio range of 1.44-7.12 and a wide range of injection timing. At fixed injection timing, an optimal swirl ratio for minimum CO emission and fuel consumption was found. At fixed swirl ratio, CO emission and fuel consumption generally decreased as injection timing was advanced. Moreover, a sudden decrease in CO emission was observed at early injection timings. Multi-dimensional numerical simulations, pressure-based measurements of ignition delay and apparent heat release, estimates of peak flame temperature, imaging of natural combustion luminosity and spray/wall interactions, and Laser Doppler Velocimeter (LDV) measurements of in-cylinder turbulence levels are employed to clarify the sources of the observed behavior.

Neural networks are useful tools for optimization studies since they are very fast, so that while capturing the accuracy of multi-dimensional CFD calculations or experimental data, they can be run numerous times as required by many optimization techniques. This paper describes how a set of neural networks trained on a multi-dimensional CFD code to predict pressure, temperature, heat flux, torque and emissions, have been used by a genetic algorithm in combination with a hill-climbing type algorithm to optimize operating parameters of a diesel engine over the entire speed-torque map of the engine. The optimized parameters are mass of fuel injected per cycle, shape of the injection profile for dual split injection, start of injection, EGR level and boost pressure. These have been optimized for minimum emissions. Another set of neural networks have been trained to predict the optimized parameters, based on the speed-torque point of the engine.

Neural networks have been used for engine computations in the recent past. One reason for using neural networks is to capture the accuracy of multi-dimensional CFD calculations or experimental data while saving computational time, so that system simulations can be performed within a reasonable time frame. This paper describes three methods to improve upon neural network predictions. Improvement is demonstrated for in-cylinder pressure predictions in particular. The first method incorporates a physical combustion model within the transfer function of the neural network, so that the network predictions incorporate physical relationships as well as mathematical models to fit the data. The second method shows how partitioning the data into different regimes based on different physical processes, and training different networks for different regimes, improves the accuracy of predictions.

Combustion under stratified operating conditions in a direct-injection spark-ignition engine was investigated using simultaneous planar laser-induced fluorescence imaging of the fuel distribution (via 3-pentanone doped into the fuel) and the combustion products (via OH, which occurs naturally). The simultaneous images allow direct determination of the flame front location under highly stratified conditions where the flame, or product, location is not uniquely identified by the absence of fuel. The 3-pentanone images were quantified, and an edge detection algorithm was developed and applied to the OH data to identify the flame front position. The result was the compilation of local flame-front equivalence ratio probability density functions (PDFs) for engine operating conditions at 600 and 1200 rpm and engine loads varying from equivalence ratios of 0.89 to 0.32 with an unthrottled intake. Homogeneous conditions were used to verify the integrity of the method.

To overcome the trade-off between NOx and particulate emissions for future diesel vehicles and engines it is necessary to seek methods to lower pollutant emissions. The desired simultaneous improvement in fuel efficiency for future DI (Direct Injection) diesels is also a difficult challenge due to the combustion modifications that will be required to meet the exhaust emission mandates. This study demonstrates the emission reduction capability of split injections, EGR (Exhaust Gas Recirculation), and other parameters on a High Speed Direct Injection (HSDI) diesel engine equipped with a common rail injection system using an RSM (Response Surface Method) optimization method. The optimizations were conducted at 1757 rev/min, 45% load. Six factors were considered for the optimization, namely the EGR rate, SOI (Start of Injection), intake boost pressure, and injection pressure, the percentage of fuel in the first injection, and the dwell between injections.

A study of the combined use of split injections, EGR, and flexible boosting was conducted. Statistical optimization of the engine operating parameters was accomplished using a new response surface method. The objective of the study was to demonstrate the emissions and fuel consumption capabilities of a state-of-the-art heavy -duty diesel engine when using split injections, EGR, and flexible boosting over a wide range of engine operating conditions. Previous studies have indicated that multiple injections with EGR can provide substantial simultaneous reductions in emissions of particulate and NOx from heavy-duty diesel engines, but careful optimization of the operating parameters is necessary in order to receive the full benefit of these combustion control techniques. Similarly, boost has been shown to be an important parameter to optimize. During the experiments, an instrumented single-cylinder heavy -duty diesel engine was used.

A study of statistical optimization of engine operating parameters was conducted. The objective of the study was to develop a strategy to efficiently optimize operating parameters of diesel engines with multiple injection and EGR capabilities. Previous studies have indicated that multiple injections with EGR can provide substantial simultaneous reductions in emissions of particulate and NOx from heavy-duty diesel engines, but careful optimization of the operating parameters is necessary in order to receive the full benefit of these combustion control techniques. The goal of the present study was to optimize the control parameters to reduce emissions and brake specific fuel consumption. An instrumented single-cylinder heavy-duty diesel engine was used with a prototype mechanically actuated (cam driven) fuel injection system.

An electronically controlled Caterpillar single-cylinder oil test engine (SCOTE) was used to study diesel combustion. The SCOTE retains the port, combustion chamber, and injection geometry of the production six cylinder, 373 kW (500 hp) 3406E heavy-duty truck engine. The engine was equipped with an electronic unit injector and an electronically controlled common rail injector that is capable of multiple injections. An emissions investigation was carried out using a six-mode cycle simulation of the EPA Federal Transient Test Procedure. The results show that the SCOTE meets current EPA mandated emissions levels, despite the higher internal friction imposed by the single-cylinder configuration. NOx versus particulate trade-off curves were generated over a range of injection timings for each mode and results of heat release calculations were examined, giving insight into combustion phenomena in current “state of the art” heavy-duty diesel engines.