The history of planetary exploration trace back to the 20th July 1969, when the first human footprint was impressed on the surface of the moon. However, exploring other planets with human crews is currently impossible to realize. Besides he technical difficulties, the main issue regards the huge distances involved and the long time required to reach such remote regions of the Solar System. For that reason, robotics and autonomous robots in particular, will play an essential role in the future of planetary exploration. Autonomy is crucial as the more a robot is far from the earth, the more it should be able to rely on its own abilities to accomplish its mission. When communication delay between the robot and he Earth is hours, devising advanced autonomous capability for an exploring robot is the only route toward the expansion of our knowledge into deep space.
Only recently, under the mission Mars Pathfinder, the first ever robotic exploration vehicle, called Sojourner, landed on the Martian surface in 1997. After Mars Pathfinder, more sophisticated robots, such as the rovers Spirit and Opportunity, were landed on Mars in 2004. The rovers were designed to withstand harsh Martian conditions for only 90 days, although after four years they are still exploring Mars and bringing new discoveries. The future NASA’s rover mission is called Mars Science Laboratory (MSL) and it is to be launched in 2009. This mission involves a rover carrying more sophisticated instruments that will help answering the questions about Mars history, climate, geology, possible life and it will also prepare for future human exploration. Alongside the NASA projects, several other projects are under development by the European Spatial Agency, as well as China and Japan. Among the several tasks that a robot devoted to explore a planet surface has to accomplish, the ability to move autonomously within an unknown environment is a basic one. In particular, such a robot must be capable of navigating in a new environment and avoiding obstacles that force the robot to deviate from its route. In addition, the obstacles can have different characteristics, such as big rocks or holes in the terrain. These differences require the robot to have the ability to distinguish between the different types of obstacles and actuate the appropriate avoidance maneuvers. The above-mentioned rovers Sojourner, Spirit and Opportunity, use stereo cameras for navigation and obstacle avoidance.
The two more recent robots Spirit and Opportunity, in particular, are equipped with three sets of stereo camera pairs. One pair is looking forward, under the solar panel in front. Another pair is looking backward, under the solar panel in he back, and the last pair is placed on the mast. This camera is mainly used for navigation purposes. With the images taken by the cameras, a stereo algorithm calculate the 3D representation of the terrain in front of the robot and other logarithms are used to calculate a “traversability” map. The information of this map is then used to calculate the next action of the robot. However, there are no other means for the rovers to sense the obstacles if these cameras failed. For this reason, it is worth to explore other possible solutions that allow the rovers to navigate and avoid obstacles, besides the use of stereo cameras. These alternative methods might represent useful complements in the sensory systems of robot which has to operate in difficult conditions into deep space, where any possible human intervention is prevented by the huge communication delays. In this paper we will explore the feasibility of an alternative obstacle avoidance system based on a set of infrared sensors that provide the robots with information about the presence of obstacles within a given range in its proximity. The system presented is able to deal with different types of objects, such as rocks and holes, and it is based on evolutionary robotics (ER) techniques. To investigate this alternative methodology, a 3D physics rover as well as a terrain model was built using Open Dynamics Engine (ODE), which is an open source library for simulating rigid body dynamics (www.ode.org). The computer model of the rover is based on the approximate dimensions of the MSL rover and its “brain”, its control system, consists of an artificial neural network (ANN) which synaptic weights were evolved using evolutionary computation techniques. This approach is commonly known as evolutionary robotics. Evolutionary robotics is inspired by the Darwinian principle of selective reproduction of the fittest and attempts to develop sensory-motor control systems for autonomous robots in an automated manner. Within the field of evolutionary robotics, obstacle avoidance and navigation behaviours are well known topics that have been widely used in the past to demonstrate the feasibility of the evolutionary approach in the robotic domain. In particular, those behaviours have been the ideal test bed used by evolutionary robotics to show the inseparable interconnection between the control system, the body and the environment in which the robot is operating. Alongside the scientific interests that often underpin the experiment in evolutionary robotics, the practical aim of this research is to extend the domain of the evolutionary techniques to the realm of planet exploration. To do that, not only we will have to evolve a control system capable of avoiding obstacles, but we need to face all the complexity of a hypothetical exploratory mission on the planetary surface, i.e. exploring an unknown environment by autonomously finding an effective route on rough surface full of obstacles in a safe mode and by taking into account the limited computational capability of the on-board hardware. The accomplishment of such a task requires, on one hand, a control system that must be able to sense the different types of obstacles and to deal with a rough terrain that often can make hard to navigate on it. That is, the robot should autonomously understand when a terrain is safe for navigation or when it is better to change direction. On the other hand, the limited on-board computing power forces us to reduce the complexity of the algorithms that provide the required navigation capabilities.
This research in developing an autonomous Mars rover started as my final year undergraduate project. The main deliverable of this project was only a 3d physical model of a Mars rover, however, things were going better than expected so we started additional work to make the rover autonomous. We have also researched different ways to tackle obstacle avoidance, which can be found in Adaptive Sensing and Active Vision subsections.