André Carmona Hernandes

A New Control Architecture for Robust Controllers in Rear Electric Traction Passenger HEVs

It is well known that control systems are the core of electronic differential systems (EDSs) in electric vehicles (EVs)/hybrid HEVs (HEVs). However, conventional closed-loop control architectures do not completely match the needed ability to reject noises/disturbances, especially regarding the input acceleration signal incoming from the drivers commands, which makes the EDS (in this case) ineffective. Due to this, in this paper, a novel EDS control architecture is proposed to offer a new approach for the traction system that can be used with a great variety of controllers (e.g., classic, artificial intelligence (AI)-based, and modern/robust theory). In addition to this, a modified proportional-integral derivative (PID) controller, an AI-based neuro-fuzzy controller, and a robust optimal H∞ controller were designed and evaluated to observe and evaluate the versatility of the novel architecture. Kinematic and dynamic models of the vehicle are briefly introduced. Then, simulated and experimental results were presented and discussed. A Hybrid Electric Vehicle in Low Scale (HELVIS)-Sim simulation environment was employed to the preliminary analysis of the proposed EDS architecture. Later, the EDS itself was embedded in a dSpace 1103 high-performance interface board so that real-time control of the rear wheels of the HELVIS platform was successfully achieved.

Vision-based system for pedestrian recognition using a tuned SVM classifier

Pedestrian recognition is one of the main advantages of the currently introduced autonomous cars. It is expected that millions of lives will be saved just by implementing this technology in real roads. We study this problem from two points of view, i.e., the recognition algorithm and the data. A trained binary classifier based on a tuned RBF-kernel SVM is used for predicting pedestrians on new scenarios. It is shown that tuning this classifier improves significantly the performance when compared with an SVM with linear kernel. The images are pre-processed using the HOG algorithm in order to get a pedestrian descriptor. The prediction is evaluated using the F1 score instead of the accuracy, because it presents a better estimation of the model performance, and yields a better way of tuning the model. The model is validated using the cross-validation method; the averaged accuracy and F1 score reached were 95% and 96.3% respectively. A database composed of 5,000 pedestrian and non-pedestrian images is used for training and testing the classifier. Several pedestrian images are analyzed after applying the algorithm to determine how the database should be completed in order to improve the detection in actual road scenarios.

A comparison between reactive potential fields and Attractor Dynamics

We study four established reactive approaches that can be implemented on computationally weak hardware with the goal of minimizing oscillatory movements to reduce the energetic cost of robot navigation. In this regard, we examine the smoothness and variability of the control action in our analysis. Sensor noise, including the large variance of GPS estimates, is evaluated. Care is taken to make the techniques comparable. Statistical data were obtained in simulation in which environments were randomly varied. We show that the second order Attractor Dynamics Approach satisfies our requirements significantly better than Potential Field approaches.

Ceiling analysis of pedestrian recognition pipeline for an autonomous car application

This paper presents an exploration of the ceiling analysis of machine learning systems. It also provides an approach to the development of pedestrian recognition systems using this analysis. A pedestrian detection pipeline is simulated in order to evaluate this method. The advantage of this method is that it allows determining the most promising pipelines elements to be modified as a way of more efficiently improving the recognition system. The pedestrian recognition is based on computer vision and is intended for an autonomous car application. A Linear SVM used as classifier enables the recognition, so this development is also addressed as a machine learning problem. This analysis concludes that for this application the more worthy path to be followed is the improvement of the pre-processing method instead of the classifier.

Novel hybrid electric motor glider-quadrotor MAV for in-flight/V-STOL launching

This work presents a novel lightweight electric UAV that features fixed-wing motor glider aircraft and quadrotor helicopter capabilities. This paper presents the hybrid concept, design, evaluation and operation of a MAV (Mini Aerial Vehicle) named Sharky, fully designed and crafted by ART (Aerial Robots Team), which may be a versatile flying robot to broaden the scope of a great number of autonomous/tele-operated missions. To illustrate, Sharky may be potentially useful on precise positioning of sensors/equipment at any point in water/ground/air areas. The MAV may aid atmospheric sensing, water sample collecting, precise positioning of sensor for agriculture, surveillance of restricted/non-structured areas, such as post-disaster sites. The aircraft is morphologically and aerodynamically shaped to perform well defined and specific features, e.g., in-flight stable launching from a carrier, gliding ability, powered flight (motor-glider), transition between glider and quadrotor (and vice versa) and base level launching either as a quadrotor or a motor glider. Sharky transition (glider/quadrotor/glider) may be achieved at anytime during the mission. The aircraft center of mass is slightly shifted to offer gliding/motor gliding stability. Because it is a quadrotor, Sharky may either work as an inverted pendulum problem. Thus, translations and rotations are easily achieved using part of the potential energy from center of mass unbalance. Still, Sharky is easily able to return back to glider/motor glider configuration by using the same principle. That helps minimizing brushless motors usage and, therefore, battery consumption. Dynamic models are presented and analyzed. Sharky stability and controllability are first evaluated in VLM/Panels software. Secondly, wind tunnel analysis are run.

GISA: A Brazilian platform for autonomous cars trials

According to WHO, traffic safety is one of the major concerns of this decade. Due to this, researchers worldwide aim to reduce the large number of fatalities in traffic accidents. This paper presents GISA as a contribution to this scenario. It consists of a platform for testing autonomous car algorithms. Firstly, some requisites to build an autonomous vehicle are presented, followed by sensors, their placement and ROS middleware. Some tests are presented to check the platform performance, including localization issues and obstacle detection. Results show that GISA is a consistent platform for implementation of autonomous car algorithms.

HiL evaluation of an on-chip-based optimal H ∞ controller on the stability of a MAV in flight simulation

This work introduces a novel methodology to assist the evaluation of a wide class of control algorithms for all-type aircrafts using Hardware-in-the-Loop (HiL) based flight simulation. The originality of this paper is using Microsoft Flight Simulator (MSFS) as the environment to embed both dynamic and graphic models of Ascending Technologies Pelican MAV (Mini Aerial Vehicle) flying robot, and using an external embedded system to control the attitude of this virtual aircraft. The full duplex communication between the virtual aircraft and the control algorithm is achieved by a custom Delphi software named FVMS (Flight Variables Management System), presented in its first version in Aeroconf 2013, fully developed by Aerial Robots Team (ART). FVMS software is able to reach (read/write) a large number of flight variables from MSFS, send them to the external hardware and receive back the control signals to actuate in the virtual aircraft. We first completely present FVMS system architecture and main features. Later, the synthesis and the application of the optimal H-Infinity robust control algorithm in the external embedded system, connected to FMVS through USB interface at 115.200 bits/s of baud rate. The microcontroller used was the Microchip 16-bits dsPIC33FJ128MC706A, which assumes the complete attitude control of the virtual Pelican MAV robot. Regarding MAVs control evaluation, HiL simulation, considerably contributes to save battery time, to ease control synthesis and prototyping and to prevent accidents during tests with the real robot. The final goal is to evaluate the stability of the Pelican platform in hovering tasks in flight simulation focusing on the efficiency of the external hardware to properly deal with the optimal H-Infinity robust control algorithm actions. The HiL control of the MAV capabilities can be extended to assist the design of other classes of controllers. Such methodology allows designers to expedite control changes, e.g., gain modification in execution time.

SquidCop: Design and evaluation of a novel quadrotor MAV for in-flight launching air-to-ground missions

Energy limitations play a major drawback in aerial robotics. Regarding the deployment of a MAV (Mini-Aerial-Vehicle), depending on the distance between base and robot, most of battery charge is simply wasted in the robot round trip. Non-structured or difficult-access locations may detract from the task or be unreachable to a flying robot in terms of energy capacity. In that sense, a novel category of MAVs that may be in-flight launched may broaden and optimize the scope and quality of several tasks not covered by current MAVs and UAVs (Unmanned Aerial Vehicle). The originality of this work is to introduce a specially shaped quadrotor named SquidCop, which may perform autonomous and stable in-flight launching from a carrier aircraft. SquidCop is a low scale MAV fully designed by ART (Aerial Robots Team), which is intended to broaden the scope and range of missions, indoor, outdoor or both together. SquidCop is aerodynamically designed to offer passive stability during release, free fall and positioning at a certain point in the space. Overall drag coefficient is carefully calculated to provide correct sink rate. Still, structural integrity is guaranteed from stress. Such features ensure correct attitude angle and minimum usage of reverse power from electric motors. Unlike regular missions where MAVs necessarily use battery power between the base and the point of interest, SquidCop is intended to use minimum battery charge at descent. Hence, more battery charge may be available to be used from the very moment when mission starts. Low-cost assembly is suitable for extreme missions that may potentially represent the complete loss of the equipment. Complete analysis involves modeling, CFD (Computational Fluid Dynamics) and wind tunnel evaluation.

Tuned Support Vector Machine Classifier for Pedestrian Recognition in Urban Traffic

La busqueda de autonomia en desplazamiento automotriz esta en auge. En dicha busqueda existen diversos problemas a resolver; uno de ellos es el reconocimiento de caracteristicas en ambientes urbanos que permitan a un vehiculo tomar decisiones autonomas. Uno de los aspectos mas relevantes corresponde al reconocimiento de transeuntes, una tecnologia que se espera pueda salvar millones de vidas en accidentes de atropellamientos. En este trabajo de investigacion se propone el reconocimiento de transeuntes en ambientes urbanos por medio de un clasificador basado en una Maquina de Vectores de Soporte ajustada; se consideraron hasta 5000 imagenes de la base de datos la INRIA con el fin de entrenar el clasificador y validar su precision por medio del metodo de validacion cruzada.

Parametric Vehicular Simulator in the Evaluation of HELVIS Mini-HEV EDS Control

HELVIS (Hybrid Electric Vehicle In Low Scale) is a mini-HEV platform used on the research of HEVs (Hybrid Electric Vehicles), through which students of all degrees have the opportunity to be introduced to the universe that surrounds HEVs in many aspects. In this work the HELVIS-Sim is presented. HELVIS-Sim is a full dynamic & kinematic vehicular simulator for the HELVIS platform, consisting of a Simulink™ environment through which the states of a large number of variables related to the vehicle can be observed and analyzed. Specially in this paper, the focus is in the control of HELVIS EDS (Electronic Differential System), presenting classic, A.I.-based (Artificial Intelligence) and optimal robust controllers in the problem of the adjustment of the rear angular speeds. HELVIS-Sim results are then compared to experimental data obtained from the real HELVIS EDS, with the aid of a dSpace™ real time interface board.Copyright