Efficient Autonomous navigation for planetary rovers with limited resources (Paper Review)-Part 1

By -

 Paper Name:- Efficient autonomous navigation for planetary rovers with limited resources

Author Name:- Levin Gerdes, Martin Azkarat, José Ricardo Sánchez‐Ibáñez, Luc Joudrier, Carlos Jesús Perez‐del‐Pulgar

DOI:- 10.1002/rob.21981
(Please refer to the paper for the reference that I have mentioned in the review and want to know more about the topic)



Introduction:-

The ExoMars rover, one of the European Space Agency's most famous missions to date, will launch in 2022. ExoMars is a six-wheeled rover with a sophisticated drill system and laboratory equipment that will study Mars' underground soil in search of life. The Trace Gas Orbiter (TGO), launched in early 2018, will serve as a communication relay, allowing bidirectional contact between the rover and the ground. Communication with the rover may only be possible twice every Sol due to limits on the allocation of deep space antennas and the TGO's orbital track (Martian day). The rover can only relay telemetry data back to Earth and receive fresh commands from ground operators during these limited connection intervals. Real-time supervised control is unfeasible because to these communication limits, as well as the signal propagation delay between Earth and Mars. As a result, it's important to focus on the rover's autonomous capabilities, which will boost the overall mission's scientific yield.

The Mars Exploration Rovers (MER) were the first rovers to have some autonomous navigational features. The earliest computer vision (CV) methods employed by MER to accomplish onboard relative localization using visual odometry (VO) are shown in Cheng, Maimone, and Matthies (2005), Maimone, Cheng, and Matthies (2007), and Matthies (2007).  Maimone (2017) provides more information on the latest features for autonomous navigation of Mars rovers. There are two major implementations of VO for space in Europe. VisLoc (Shaw, Woods, Churchill, & Newman, 2013), which was developed as part of the ExoMars mission and is based on the Oxford visual odometry (OVO) algorithm (Churchill & Newman, 2012), and SPARTAN (Kostavelis, Boukas, Nalpantidis, Gasteratos, & AvilesRodrigalvarez, 2011, 2014), which was developed as part of ESA R&D activities during the last decade. Furthermore, recent SPARTAN work has focused on improving and bringing localization and mapping procedures to space representative FPGA hardware implementations (Avilés 2018; Lentaris, Maragos, Soudris, Zabulis, & Lourakis, 2019). To improve the technical readiness level (TRL) of these technologies, field test programs similar to ExoMars have been conducted in the Atacama Desert. The rover performed an independent traverse of 5.05 kilometers in one day during the SEEKER campaign (Woods, 2014). Additionally, the French National Centre for Space Studies (CNES) has developed AutoNav (Bousquet, 2011; Moreno, 2013), an autonomous navigation solution that has been demonstrated to work in both simulation and shorter field operations. Throughout the development of the ExoMars rover, the industry has presented designs and developments for autonomous navigation (Bora, Nye, Lancaster, Barclay, & Winter, 2017; Winter, 2015, 2017), which will eventually be implemented in the rover (Bora, Nye, Lancaster, Barclay, & Winter, 2017; Winter, 2015, 2017). These functionalities, however, have only been explored in simulated environments so far. Overall, despite the fact that the European industry and ESA have been working on the development of several autonomous navigation capabilities for many years, the ExoMars rover onboard flight SW version containing autonomous navigation functionality is not yet ready, and only a version with a navigation mode based on the path following approach has been validated, which contains the VisLoc algorithm flight code, due to project budget and schedule constraints (Townson, Woods, & Carnochan, 2018). It is understood that efficiency and computational resources are known restrictions for these functionalities in the process of creating autonomous navigation features. Due to the required frequent stops and analysis of sensor data to analyze the terrain before executing the course ahead, it is expected that ExoMars autonomous navigation mode will drastically limit the effective rover average speed. To build a map, the poor computational onboard capabilities will need tens of seconds, and to perform a self-localization loop, it will take on the order of second(s). In a manner similar to the navigation approach suggested by Correal and Pajares, both the localization and mapping processes must run sequentially and in turn, such that localization runs while driving and mapping runs while halted (2011).

With the more basic path following navigation baselined for ExoMars, the rover will be handled similarly to NASA's MER and Curiosity rovers in their early days. A full list of duties defined by a team of specialists will be sent to the rover for each Sol. The duties will involve following a path that has been defined on the ground in the form of navigation waypoints for the rover to follow. Because the rover will not be able to detect any obstructions along the path, these locations must make a safe traversable path. The constraint in this scenario comes from the rover's telemetry data, which the operators use to judge the terrain's traversability and estimate dangers, mostly from the Navigation Cameras stereo bench, rather than the average rover traverse speed. The longer the trip, the more challenging it is to ensure safe navigation due to potential occlusions, decreased stereo quality at greater distances, and drift in localization over travelled distance. Because of the requirement for conservative planning, the prescribed distance to move every Sol may not utilise all of the energy available to the locomotion system, limiting the mission science return.

The authors of this work propose using computationally efficient safety features to compensate for autonomous navigation capabilities that are either absent or computationally expensive. In the case of ExoMars, these qualities could allow for less cautious planning, allowing for longer scheduled traverses per Sol while still maintaining the rover's safety, when compared to its path following navigation mode. Reactive hazard detection is one of the proposed characteristics. With a rover that can dependably detect risks and stop on its own, the operators can command longer traverses with more confidence. The authors also look into how the daily travelled path length might be enhanced by allowing dynamic onboard path replanning for hazard avoidance as a second feature. Instead of pausing in front of a hazard, we propose that the rover registers the obstacle, determines its position relative to the rover, and then navigates around the obstruction before returning to the original course. As mentioned in the title of the paper, the author could describe this navigation mode as a path following with reactive hazard avoidance, and we refer to it as efficient autonomous navigation. The proposed characteristics give a higher level of autonomy than the path-following technique, but they still fall short of fully autonomous navigation because they do not include a comprehensive terrain mapping and traversability analysis process. Nonetheless, depending on the travelled terrain hardness, these traits may equip the rover with sufficient skills to autonomously go to a target in an efficient and relatively quick manner.

It's worth noting that, compared to the default implementation, these features have a negligible processing overhead. To begin with, the hazard detection function is based solely on stereo disparity values computed across the localization cameras, which are already used to parts of the camera picture frames for VO estimations and hence need very little additional computing time. Second, the danger avoidance function only has to operate at discrete events, which has a low impact on total processing resources. To provide this capability, however, the current onboard capabilities would need to be enhanced. We highlight efficiency as an essential motivation for our work because of the aforementioned resource restrictions and constraints of space missions.

Note:- We will discuss this paper further in the next blog, especially on the building block of efficient navigation.


Comments

Post a Comment

Please do not use foul language while commenting. We will appreciate your suggestions and reviews in the comments.

Read Also

Deep Neural Networks for ADMET properties' prediction

By -
Image
Introduction The absorption, distribution, metabolism, excretion and toxicity properties of any compound contributes to its potential as an orally administered drug. A set of criteria defined by Rule of Five or the Pfizer's Rule of Five stats that any compound meeting these criteria are more likely to have ADMET properties. The importance of these properties has increased in modern medicinal chemistry and new machine learning techniques are employed to predict these properties. The predictive ADMET models have helped in discovery of small molecules with improved safety and dose. Deep Neural Networks are showing usefulness in such predictions due to improved computational efficiency, larger datasets and adaption of image processing in the chemistry. Early Models The early DNNs employed showed improved prediction performance. Many researchers raised concerned that this improvement is in fact has no relation with better computational capacity, rather it is result of mere memorizing of
Read more

Marine eDNA Analysis using DL techniques

By -
Image
Introduction Environmental DNA (eDNA) is the genetic information or material released by an organism into its surroundings through activities like skin shedding, mucus, urine or other secretions. These materials contain DNA fragments that can be found in soil, water or air samples. The concept of eDNA is based on the fact that organisms constantly leave organic traces behind them. By analysing the eDNA, researchers can detect the presence of any species in that area. This method is a valuable technique for biodiversity monitoring and ecological research.    Unlike traditional methods that involve direct observation or capturing of organisms, eDNA analysis allows researchers to study biodiversity without directly interacting with the species. This non-invasive approach is particularly useful for studying elusive or endangered species. eDNA analysis has been applied in various ecosystems, including freshwater, marine, and terrestrial environments. It is used to study aquatic organisms, t
Read more

Assisting Neuroimaging through DL

By -
Image
Introduction: In recent years, neuroimaging has been a research attraction, especially due to fMRI (Functional Magnetic Resonance Imaging) advancements. The fMRI data comprises of often several individuals. This data is used by researchers to study the association between the cognitive states of an individual and the underlying brain activities. The fMRI data used for neuroimaging is ideally suited for deep learning applications. The large, structured datasets of fMRI can be used for representation-learning methods of the DL. Generally, DL can be defined as a learning method with multiple levels of abstraction, where at each level the input data is transformed by a simple non-linear function which enables the model to recognize complex patterns. With higher-level representation, DL methods can associate a target variable with variable patterns in input data. Also, DL techniques can independently acquire these transformations from the data, eliminating the need for a comprehensive preex
Read more

What is Terraforming?

By -
Image
  Humans are working more and more on extraterrestrial habitations. A few decades back we were planning to settle on the Moon and now we are planning the same on Mars. Many researchers also proposed many theories on such habitations as O'Neill Cylinder, McKendree Cylinder, and many more. One such famous hypothetical theory is Terraforming. But what does the term terraforming mean? Terraforming is the process of transforming the atmosphere, topography, temperature and other factors of any moon or planet to make it habitable and similar to the Earth. It was Carl Sagan who first proposed the theory of terraforming. He proposed the terraforming of Venus. He thought of seeding Venus with algae. The algae will convert nitrogen and carbon dioxide to organic matter and eventually, the greenhouse effect will reduce to a comfortable, habitual level.  However, later studies revealed that the clouds of Venus are full of sulphuric acid. Also, the atmosphere of Venus is very thick, any carbon co
Read more