Figure - uploaded by Antonio Mora
Content may be subject to copyright.
Source publication
When driving a car it is essential to take into account all possible factors; even more so when, like in the TORCS simulated race game, the objective is not only to avoid collisions, but also to win the race within a limited budget. In this paper, we present the design of an autonomous driver for racing car in a simulated race. Unlike previous cont...
Context in source publication
Similar publications
At the beginning of the article the authors describe a new trend in the artificial intelligence, associated with fuzzy sets and the accompanying derivative solutions which may include L-R numbers by Dubois and Prade. On their basis the redefined theory has started a new trend in the form of ordered fuzzy numbers (OFN). Main features of ordered fuzz...
A number of similarities between fuzzy logic and probability can be identified. Historically, fuzzy logic is founded upon a probability based many valued logic of Lukasiewicz. The fuzzy membership function and the probability density function closely resemble as they both have similar shapes and have values that range between zero and one. Vague li...
This paper presents the design of a speed control for a DC motor using fuzzy logic by software LabView, is also a literature review the design and implementation environment is presented by fuzzy logic describing the materials and methods used. Various processes on the subject highlighting the idea, creation, development and implementation of intel...
This paper presents a new control topology to reduce the effect of the fifth negative and seventh positive voltages harmonic in the Doubly Fed Induction Generator (DFIG) in wind power systems. The proposed control is based on the concept of active power filters. A rotor current is delivered in the opposite direction of the harmonic content to reduc...
One of the biomedical image problems is the appearance of the bubbles in the slide that could occur when air passes through the slide during the preparation process. These bubbles may complicate the process of analysing the histopathological images. Aims: The objective of this study is to remove the bubble noise from the histopathology images, and...
Citations
... Although the car is driving successfully around the racetrack and follows a raceline the controller is only able to handle low speeds. In both [164], [166] the authors propose two fuzzy controllers for calculating the steering angle and computing the target speed of the car based on sensor information in the TORCS simulator [241]. In Oliveira et al. [142] Bayesian optimization (BO) is used to find a control policy that minimizes the time per lap while keeping the vehicle on the racetrack. ...
The rising popularity of self-driving cars has led to the emergence of a new research field in the recent years: Autonomous racing. Researchers are developing software and hardware for high performance race vehicles which aim to operate autonomously on the edge of the vehicles limits: High speeds, high accelerations, low reaction times, highly uncertain, dynamic and adversarial environments. This paper represents the first holistic survey that covers the research in the field of autonomous racing. We focus on the field of autonomous racecars only and display the algorithms, methods and approaches that are used in the fields of perception, planning and control as well as end-to-end learning. Further, with an increasing number of autonomous racing competitions, researchers now have access to a range of high performance platforms to test and evaluate their autonomy algorithms. This survey presents a comprehensive overview of the current autonomous racing platforms emphasizing both the software-hardware co-evolution to the current stage. Finally, based on additional discussion with leading researchers in the field we conclude with a summary of open research challenges that will guide future researchers in this field.
... Although the car is driving successfully around the racetrack and follows a raceline the controller is only able to handle low speeds. In both [164], [166] the authors propose two fuzzy controllers for calculating the steering angle and computing the target speed of the car based on sensor information in the TORCS simulator [241]. In Oliveira et al. [142] Bayesian optimization (BO) is used to find a control policy that minimizes the time per lap while keeping the vehicle on the racetrack. ...
The rising popularity of self-driving cars has led to the emergence of a new research field inrecent years: Autonomous racing. Researchers are developing software and hardware for high-performancerace vehicles which aim to operate autonomously on the edge of the vehicle’s limits: High speeds, highaccelerations, low reaction times, highly uncertain, dynamic, and adversarial environments. This paperrepresents the first holistic survey that covers the research in the field of autonomous racing. We focuson the field of autonomous racecars only and display the algorithms, methods, and approaches used inthe areas of perception, planning, control, and end-to-end learning. Further, with an increasing numberof autonomous racing competitions, researchers now have access to high-performance platforms to testand evaluate their autonomy algorithms. This survey presents a comprehensive overview of the currentautonomous racing platforms, emphasizing the software-hardware co-evolution to the current stage. Finally,based on additional discussion with leading researchers in the field, we conclude with a summary of openresearch challenges that will guide future researchers in this field.
... At the symbolic level, NUTs is most similar to fuzzy decision system [26], a representation that has actually been used before in AV context, specifically as a TORCS driving controller [32]. There are subtle differences however. ...
A probabilistic graphical model, Neuronal Unit of Thoughts (NUTs), is proposed in this paper that offers a formalism for the integration of lower-level cognitions. Nodes or neurons in NUTs represent sensory data or mental concepts or actions, and edges the causal relation between them. A node affects a change in the Action Potential(AP) of its child node, triggering a value change once the AP reaches a fuzzy threshold. Multiple NUTs may be crossed together producing a novel NUTs. The transition time in a NUTs, in response to a 'surprise', is characterized , and the formalism is evaluated in the context of a non-trivial application: Autonomous Driving with imperfect sensors.
... Previously, the authors presented an approach combining two specialized fuzzy controllers, designed by hand, that were able to decide the car's proper steering angle and desired speed at every single point (or tick) during a race [29]. This driver was later improved [30] optimizing the parameters of their membership functions by means of a Genetic Algorithm [8]; this automated design improved manual one obtaining several controllers that were able to beat the initial hand-designed controller in a race, as well as other published controllers. ...
... We initially proposed a controller [29] with the same modular architecture as the simple TORCS driver; however, the target speed and steering angle are computed by means of two modular and specialized fuzzy sub-controllers, which consider five position sensors. This is the controller which will be improved by means of a GA in this work. ...
... Moreover, the controller is based in a set of fuzzy rules, designed to maximize the car speed depending on the distance to the track border. These rules can be consulted in [29]. The second is the fuzzy steering sub-controller, which aims to define the steer angle estimating and determining the target position of the car. ...
... A fuzzy controller works with linguistic variables, and will for instance turn slightly to the right when the next curve is close, but these controllers have to be designed to map properly inputs to desired outputs in particular situations. Previously, the authors presented an approach combining two specialized fuzzy controllers, designed by hand, that were able to decide the car's proper steering angle and desired speed at every single point (or tick) during a race [26]. This driver was later improved [27] optimizing the parameters of their membership functions by means of a Genetic Algorithm [7]; this automated design improved manual one obtaining several controllers that were able to beat the initial hand-designed controller in a race, as well as other published controllers. ...
... Onieva et al. [22] presented a parametrized modular architecture with a fuzzy system and a GA, aiming to reproduce the behavior of a human driver when controlling the steering wheel. In this line, we presented an improved approach [26], which evolved a fuzzy-based driver considering the target speed in addition to the steer (two fuzzy sub-controllers). This controller was also enhanced in a further work [27] optimizing the parameters of the membership functions by means of a real coded genetic algorithm, obtaining a noticeable improvement in performance. ...
... The controller proposed initially [26] has the same modular architecture as the simple TORCS driver; however, the target speed and steering angle are computed by means of two modular and specialized fuzzy sub-controllers, which consider five position sensors. This is the controller which will be improved by means of a GA in this work. ...
... Among the different types of controllers that have been used in TORCS, fuzzy-based ones have proved to have one of the best performances, since they simulate in part the human reasoning when driving [20,13,19]. Thus, the authors presented previously an approach in which two specialized fuzzy controllers were combined to decide the car's steering angle and desired speed in every single point (or tick) during a race [22]. It yielded good results in several tracks, but the autonomous driver had some troubles in difficult circuits, such as those with many curves or competing against tough rivals. ...
... Onieva et al. [17] presented a parametrized modular architecture with a fuzzy system and a GA in the design of fuzzy logic controllers for steering wheel management that can reproduce human driver behavior, but it did not take the target speed into account, unlike our previous controller [22] which computed the target speed and the steer with two fuzzy sub-controllers and whose membership functions parameters were defined by trial/error process. In this paper, we propose to optimize these parameters using a real coded genetic algorithm aiming to improve the performance of the original fuzzy controller. ...
... The initial proposed controller [22] has the same modular architecture as the simple TORCS driver, however, the target speed and steering angle are computed by means of two modular and specialized fuzzy sub-controllers, which consider five position sensors. This is the controller which will be improved by means of a GA in this work. ...
... In Section 5 we generate controllers for cars in The Open Racing Car Simulator (TORCS) (Wymann et al., 2014). TORCS has been used extensively in AI research, for example in (Salem et al., 2017), (Koutník et al., 2013), and (Loiacono et al., 2010) among others. (Lillicrap et al., 2015a) has shown that a Deep Deterministic Policy Gradient (DDPG) network can be used in RL environments with continuous action spaces. ...
We study the problem of generating interpretable and verifiable policies through reinforcement learning. Unlike the popular Deep Reinforcement Learning (DRL) paradigm, in which the policy is represented by a neural network, the aim in Programmatically Interpretable Reinforcement Learning is to find a policy that can be represented in a high-level programming language. Such programmatic policies have the benefits of being more easily interpreted than neural networks, and being amenable to verification by symbolic methods. We propose a new method, called Neurally Directed Program Search (NDPS), for solving the challenging nonsmooth optimization problem of finding a programmatic policy with maxima reward. NDPS works by first learning a neural policy network using DRL, and then performing a local search over programmatic policies that seeks to minimize a distance from this neural "oracle". We evaluate NDPS on the task of learning to drive a simulated car in the TORCS car-racing environment. We demonstrate that NDPS is able to discover human-readable policies that pass some significant performance bars. We also find that a well-designed policy language can serve as a regularizer, and result in the discovery of policies that lead to smoother trajectories and are more easily transferred to environments not encountered during training.
... Many kinds of controllers have been used in this simulator, but one of the most efficient controllers so far are fuzzy-based ones, as they simulate in part the human reasoning when driving [14,19,20]. In this line, the authors presented previously an approach in which two specialized fuzzy controllers were combined to decide the car's steering angle and desired speed in every single point (or tick) during a race [22]. The obtained results were promising, but the performance of the autonomous driver showed some flaws in difficult tracks -those with many curves, or where 'external' factors affected the asphalt -and against the most competitive rivals. ...
... Onieva et al. [18] presented a parametrized modular architecture with a fuzzy system and a GA in the design of fuzzy logic controllers for steering wheel management that can reproduce human driver behavior, but it did not take the target speed into account, unlike our previous controller [22] which computed the target speed and the steer with two fuzzy sub-controllers and whose membership functions parameters were defined by trial/error process. In this paper, we propose to optimize these parameters using a real coded genetic algorithm aiming to improve the performance of the original fuzzy controller. ...
... The initial proposed controller [22] has the same modular architecture as the simple driver of TORCS, however, the target speed and steering angle are computed by means of two modular and specialized fuzzy sub-controllers, which consider five position sensors. This is the controller which will be improved by means of a GA in this work. ...
This work presents an evolutionary approach to optimize the parameters of a Fuzzy-based autonomous driver for the open simulated car racing game (TORCS). Using evolutionary algorithms, we intend to optimize a modular fuzzy agent designed to determine the optimal target speed as well as the steering angle during the race. The challenge in this kind of fuzzy systems is the design of the membership functions, which is usually done through a trial and error process, but in this paper an adapted real-coded Genetic Algorithm with two different fitness functions - has been applied to find the best values for these parameters, obtaining a robust design for the TORCS controller. The evolved drivers were tested and evaluated competing against other TORCS controllers in practice mode, without rivals, and real races. The optimized fuzzy-controllers yield a very good performance, mainly in tracks that have many turning points, which are, in turn, the most difficult for any autonomous agent. Thus, this is a real enhancement of the baseline fuzzy controllers which had several difficulties to drive in this kind of circuits.
