Malte Probst

Tim Puphal, Malte Probst, Julian Eggert

Risk assessment is a central element for the development and validation of Autonomous Vehicles (AV). It comprises a combination of occurrence probability and severity of future critical events. Time Headway (TH) as well as Time-To-Contact (TTC) are commonly used risk metrics and have qualitative relations to occurrence probability. However, they lack theoretical derivations and additionally they are designed to only cover special types of traffic scenarios (e.g. following between single car pairs). In this paper, we present a probabilistic situation risk model based on survival analysis considerations and extend it to naturally incorporate sensory, temporal and behavioral uncertainties as they arise in real-world scenarios. The resulting Risk Spot Detector (RSD) is applied and tested on naturalistic driving data of a multi-lane boulevard with several intersections, enabling the visualization of road criticality maps. Compared to TH and TTC, our approach is more selective and specific in predicting risk. RSD concentrates on driving sections of high vehicle density where large accelerations and decelerations or approaches with high velocity occur.

Tim Puphal, Malte Probst, Yiyang Li et al.

We consider the problem of correct motion planning for T-intersection merge-ins of arbitrary geometry and vehicle density. A merge-in support system has to estimate the chances that a gap between two consecutive vehicles can be taken successfully. In contrast to previous models based on heuristic gap size rules, we present an approach which optimizes the integral risk of the situation using parametrized velocity ramps. It accounts for the risks from curves and all involved vehicles (front and rear on all paths) with a so-called survival analysis. For comparison, we also introduce a specially designed extension of the Intelligent Driver Model (IDM) for entering intersections. We show in a quantitative statistical evaluation that the survival method provides advantages in terms of lower absolute risk (i.e., no crash happens) and better risk-utility tradeoff (i.e., making better use of appearing gaps). Furthermore, our approach generalizes to more complex situations with additional risk sources.

Thomas Schnürer, Malte Probst, Horst-Michael Gross

Reinforcement Learning (RL) can enable agents to learn complex tasks. However, it is difficult to interpret the knowledge and reuse it across tasks. Inductive biases can address such issues by explicitly providing generic yet useful decomposition that is otherwise difficult or expensive to learn implicitly. For example, object-centered approaches decompose a high dimensional observation into individual objects. Expanding on this, we utilize an inductive bias for explicit object-centered knowledge separation that provides further decomposition into semantic representations and dynamics knowledge. For this, we introduce a semantic module that predicts an objects' semantic state based on its context. The resulting affordance-like object state can then be used to enrich perceptual object representations. With a minimal setup and an environment that enables puzzle-like tasks, we demonstrate the feasibility and benefits of this approach. Specifically, we compare three different methods of integrating semantic representations into a model-based RL architecture. Our experiments show that the degree of explicitness in knowledge separation correlates with faster learning, better accuracy, better generalization, and better interpretability.

Tim Puphal, Raphael Wenzel, Benedict Flade et al.

Self-driving cars face complex driving situations with a large amount of agents when moving in crowded cities. However, some of the agents are actually not influencing the behavior of the self-driving car. Filtering out unimportant agents would inherently simplify the behavior or motion planning task for the system. The planning system can then focus on fewer agents to find optimal behavior solutions for the ego~agent. This is helpful especially in terms of computational efficiency. In this paper, therefore, the research topic of importance filtering with driving risk models is introduced. We give an overview of state-of-the-art risk models and present newly adapted risk models for filtering. Their capability to filter out surrounding unimportant agents is compared in a large-scale experiment. As it turns out, the novel trajectory distance balances performance, robustness and efficiency well. Based on the results, we can further derive a novel filter architecture with multiple filter steps, for which risk models are recommended for each step, to further improve the robustness. We are confident that this will enable current behavior planning systems to better solve complex situations in everyday driving.

Tim Puphal, Ryohei Hirano, Malte Probst et al.

Driver support systems that include human states in the support process is an active research field. Many recent approaches allow, for example, to sense the driver's drowsiness or awareness of the driving situation. However, so far, this rich information has not been utilized much for improving the effectiveness of support systems. In this paper, we therefore propose a warning system that uses human states in the form of driver errors and can warn users in some cases of upcoming risks several seconds earlier than the state of the art systems not considering human factors. The system consists of a behavior planner Risk Maps which directly changes its prediction of the surrounding driving situation based on the sensed driver errors. By checking if this driver's behavior plan is objectively safe, a more robust and foresighted driver warning is achieved. In different simulations of a dynamic lane change and intersection scenarios, we show how the driver's behavior plan can become unsafe, given the estimate of driver errors, and experimentally validate the advantages of considering human factors.

Malte Probst

Estimation of Distribution Algorithms (EDAs) require flexible probability models that can be efficiently learned and sampled. Generative Adversarial Networks (GAN) are generative neural networks which can be trained to implicitly model the probability distribution of given data, and it is possible to sample this distribution. We integrate a GAN into an EDA and evaluate the performance of this system when solving combinatorial optimization problems with a single objective. We use several standard benchmark problems and compare the results to state-of-the-art multivariate EDAs. GAN-EDA doe not yield competitive results - the GAN lacks the ability to quickly learn a good approximation of the probability distribution. A key reason seems to be the large amount of noise present in the first EDA generations.

Malte Probst, Franz Rothlauf

Estimation of Distribution Algorithms (EDAs) require flexible probability models that can be efficiently learned and sampled. Deep Boltzmann Machines (DBMs) are generative neural networks with these desired properties. We integrate a DBM into an EDA and evaluate the performance of this system in solving combinatorial optimization problems with a single objective. We compare the results to the Bayesian Optimization Algorithm. The performance of DBM-EDA was superior to BOA for difficult additively decomposable functions, i.e., concatenated deceptive traps of higher order. For most other benchmark problems, DBM-EDA cannot clearly outperform BOA, or other neural network-based EDAs. In particular, it often yields optimal solutions for a subset of the runs (with fewer evaluations than BOA), but is unable to provide reliable convergence to the global optimum competitively. At the same time, the model building process is computationally more expensive than that of other EDAs using probabilistic models from the neural network family, such as DAE-EDA.

Malte Probst, Franz Rothlauf

We propose an alternative method for training a classification model. Using the MNIST set of handwritten digits and Restricted Boltzmann Machines, it is possible to reach a classification performance competitive to semi-supervised learning if we first train a model in an unsupervised fashion on unlabeled data only, and then manually add labels to model samples instead of training data samples with the help of a GUI. This approach can benefit from the fact that model samples can be presented to the human labeler in a video-like fashion, resulting in a higher number of labeled examples. Also, after some initial training, hard-to-classify examples can be distinguished from easy ones automatically, saving manual work.

Malte Probst

Estimation of Distribution Algorithms (EDAs) require flexible probability models that can be efficiently learned and sampled. Autoencoders (AE) are generative stochastic networks with these desired properties. We integrate a special type of AE, the Denoising Autoencoder (DAE), into an EDA and evaluate the performance of DAE-EDA on several combinatorial optimization problems with a single objective. We asses the number of fitness evaluations as well as the required CPU times. We compare the results to the performance to the Bayesian Optimization Algorithm (BOA) and RBM-EDA, another EDA which is based on a generative neural network which has proven competitive with BOA. For the considered problem instances, DAE-EDA is considerably faster than BOA and RBM-EDA, sometimes by orders of magnitude. The number of fitness evaluations is higher than for BOA, but competitive with RBM-EDA. These results show that DAEs can be useful tools for problems with low but non-negligible fitness evaluation costs.

Malte Probst, Franz Rothlauf, Jörn Grahl

Estimation of Distribution Algorithms (EDAs) require flexible probability models that can be efficiently learned and sampled. Restricted Boltzmann Machines (RBMs) are generative neural networks with these desired properties. We integrate an RBM into an EDA and evaluate the performance of this system in solving combinatorial optimization problems with a single objective. We assess how the number of fitness evaluations and the CPU time scale with problem size and with problem complexity. The results are compared to the Bayesian Optimization Algorithm, a state-of-the-art EDA. Although RBM-EDA requires larger population sizes and a larger number of fitness evaluations, it outperforms BOA in terms of CPU times, in particular if the problem is large or complex. RBM-EDA requires less time for model building than BOA. These results highlight the potential of using generative neural networks for combinatorial optimization.