Exploring the Use of Terrestrial Robots for Atmospheric Electricity Measurement

Measuring the electric field is a central goal in electrostatics research, such as the study of electrostatics in atmospheric processes. It serves as a key indicator for various atmospheric phenomena, including the presence of lightning, dust or charged clouds. Traditionally, electric field measurements have been conducted from static platforms, with limited mobile measurements from airborne platforms such as balloons or, more recently, Unmanned Aerial Vehicles (UAVs). Here, we explore the potential of terrestrial robots to measure the electric field with some level of autonomy, such as during supervised navigation between user-defined waypoints. We mount a field mill on a four-wheeled rover and use a Global Navigation Satellite System (GNSS) to track the position of its measurements during an outdoor survey. The robot has a depth camera for 3D terrain mapping to contextualise local field measurements. We present plans for future research, including the use of semi-autonomous identification and exploration of electrostatic ‘hotspots’; and deployment of multiple robots (e.g., six) in a ‘sparse swarm’ configuration. We consider opportunities to employ such a robot system in environmental science or space research, for instance on Mars or the moon, where an understanding of electrostatic processes could be significant for future space missions.


Introduction
Measurement of the electric field is a central activity in electrostatics research, and new technological developments are opening up new methodological possibilities.While mobile platforms such as balloons or rockets, or piloted planes, have been well-employed over many years [1,2,3], autonomous mobile platforms such as unmanned aerial vehicles (UAVs) are seeing increased adoption for environmental science generally [4] and electrostatic manipulation [5] and measurement specifically [3].Although such adoption is proceeding apace in the aerial domain, relatively little electrostatics measurement research has been done with autonomous terrestrial vehicles (see [6] by Aplin and Xiong for what we believe is the first example).Such ground-based vehicles could have the advantage of longer-term, autonomous operation in a manner where UAVs might face limitations, by for example having longer battery life.A ground vehicle would be able to make stationary, in situ measurements and preserve energy, and may require less intensive supervision from a technical, safety or regulatory perspective.This might especially be the case when multiple robots are deployed simultaneously [7].Moreover, multiple measurement vehicles, in the form of a multi-robot system, may bring special advantages for understanding electrostatic phenomena that occur across wide spatial scales.This is because multiple readings could be obtained concurrently, and the sensor mobility would allow the robot system to opportunistically track or investigate electrostatic phenomena.
As well as Earth-based measurements, one domain we are especially mindful of is space exploration and the growing role of robotic systems [8], especially multi-robot systems [9,10,11].Understanding the electrostatic environment of other planets may be crucial for further space missions, including the discovery and understanding of the physical properties of planetary bodies and implications for new machinery and astronaut health.In this paper, we propose the use of a 'sparse swarm' of rovers to autonomously map and monitor the electrostatic environment.A robot swarm is characterised by decentralised organisation, such that robots have individual autonomy to react to their local environment rather than top-down control [12].There is a growing interest in the potential of swarms for planetary exploration (e.g.[13]).However, much of the current swarm robotics research assumes the use of a large number of agents, which is perhaps not feasible for space exploration.Our approach is designed with realism in mind and uses a small number of rovers in a 'sparse swarm' configuration.In a sparse swarm, there are fewer robots, much lower robot density (spatial separation), and more sporadic communication [14].This is much more realistic for practical environmental science and space research especially.
The rovers we use are 4-wheeled and capable of traversing uneven terrain and measuring the electric field (see Section 3.1).This paper begins with a brief review of current research in this field.We then describe the hardware and software of the rover, before presenting the results of some initial real-world trials with one rover.

Background
The study of atmospheric electricity both on Earth and other planets is expanding field of research.A survey of electrostatics is made by Aplin [15] which covers a comprehensive overview of the electrifying atmosphere, including charging, ionisation, and lightning in the solar system.Gao and Chien (2017) [16] provide an overview of space robotics as an emerging field of study, and our work is presented in this exciting context.The use of robotic platforms for scientific and environmental monitoring has developed significantly in the past 20 years or so.Dunbabin and Marques [17] showed the significant advantages and applications of using robots for environmental monitoring; while Katz et al. [18] show a sensing framework for 3D mapping in outdoor environments, which is closely related to our research.Schön et al. [19] have also demonstrated use of a small UAV (Unmanned Aerial Vehicle) to measure fair-weather atmospheric charge.The use of electric fields for autonomous navigation has also been explored and demonstrated by Li et al. [3].These studies show insights into the potential applications of our current research.Aplin and Xiong [6] showed the plausibility of using a terrestrial robotic platform for electric field measurement.
The longer-term goal of our research is the deployment of a 'sparse swarm' configuration of robots for electric field measurement; this concept has been explored by Tarapore et al. [14].Their work aims to move theoretical research of robot swarms into more real-world applications.Altogether, our research is done in the context of increasing use of robots for environmental monitoring, though with relatively limited use of multi-robot systems and only the earliest attempts at deploying terrestrial robots for electrostatic measurement.

Robot Hardware & Software
We present our initial configuration of a single robot for electrostatic measurement, as we work toward developing a sparse swarm system.We present the hardware and software elements.

Robot Hardware
The base robot is a 'Leo Rover' [20], a 4-wheeled robotic platform designed for Mars-like environments, which has the ability to traverse rough outdoor terrain.The base unit has a weight of 6.5 kg, dimensions of 447 × 433 × 249 mm, and a payload capacity of ca. 5 kg.Onto this base platform, we add a depth camera (Intel Realsense D455) and LiDAR (RPLIDAR-A1), both of which are critical for mapping the environment and navigation around obstacles.The depth camera is mounted on a pan-tilt unit (PhantomX XL430, Trossen Robotics) such that the camera can obtain a wide view of the surrounding area without the robot needing to move.Depth camera data will be used later for creating a 3D map of the electrostatic environment.The rover also has a Global Navigation Satellite System (GNSS) module (u-blox ZED-F9P), which allows for precise geolocation (±1 m) when recording electrostatic measurements.To acquire said electrostatic measurements the robot has an upward-facing JCI-140 electrostatic field mill attached to the back, which is directly connected to a custom-made circuit board to read the positive and negative voltages that relate to the electrostatic field.In addition to the robot's core Raspberry Pi 4B computer (2GB RAM), we add a LattePanda (3 Delta) single-board computer (SBC).This is a compact and powerful computer (2.0-2.9GHz quad-core CPU, 8GB RAM, Intel UHD Graphics) that handles all major data processing and control tasks.These hardware components can be seen in Figure 1.
3.1.1.Field mill signal An issue in our hardware setup was the need to convert the voltage output from the JCI-140 field mill into a digital signal that could be read by the onboard LattePanda.The field mill's output ranges from −2 to +2 volts, which is an issue as the onboard analogue to digital converter (ADC) of the LattePanda cannot process negative values and also has a low resolution of only 10 bits.To solve this issue, we designed a Printed Circuit Board (PCB), its schematic can be seen in Figure 2. It converts the signal from the field mill into a positive value, inverting the signal if it is negative.It also outputs an indicator voltage, which is positive if the incoming signal is positive, and 0 volts if it is negative.These two outputs are fed into a 12-bit ADC which can be read by the LattePanda.This allows the LattePanda to process the field mill's bipolar analogue signal.

Robot Grounding
To prepare the Leo Rover for testing we ensure the whole robot was grounded.This step was important as initial testing without extensive grounding indicated issues related to charges accumulating on the robot due to the motors and a possible triboelectric effect from the wheels.As the Leo Rover is not entirely made up of metal parts and has 3D-printed plastic components, we used copper tape to electrically connect all the metal parts together.We also attached a grounding chain to the back of the rover, which had sufficient length to ensure an ongoing contact with the ground during movement (Fig. 1).

Robot Software
The LattePanda runs Ubuntu 20.04 Linux and Robot Operating System (ROS Noetic).ROS is a framework for writing and using software dedicated to robotic tasks, and it has a wide range of available packages and libraries.One of the main packages that we are testing for our research is 'RTAB-Map' [21].RTAB-Map allows the robot to use its 3D camera and LiDAR sensors to do effective visual SLAM (simultaneous localisation and mapping).It builds a 3D model of its environment and tracks the robot within it.This is crucial for autonomous navigation and allows electrostatic readings to be contextualised to their local terrain.In addition to SLAM, ROS also is used to interface other sensors such as an inertial measurement unit (IMU), wheel encoders (tracking the rotation of the wheels), and GNSS location, into the ROS 'robot localization' package.This provides nonlinear state (pose) estimation through sensor fusion of an arbitrary number of sensors.We have also created our own ROS packages to record the electrostatic sensor readings and log data.In future trials we will employ the ROS 'Navigation Stack'1 to achieve autonomous navigation between waypoints that are user defined, or autonomously generated via algorithms operating in real time on the emergent, dynamic electrostatic map.

Electrostatic Modelling
Understanding the effect of the Leo Rover's geometry on the sensor readings is a crucial part of measuring the electrostatic environment accurately.We can use an electrostatic model to account for any distortion and improve the accuracy of our sensor readings.There are two main approaches to getting this information.The first way is to use software modelling.This involves creating a 3D model of the Leo Rover and simulating an electrostatic field around it.An alternative approach is in the real world to create a fixed calibration platform where the sensors are exposed to a known electric field, which can then be used to measure the robot's effect on the electrostatic readings.
For our case, we chose to start with the software approach, with plans to use a real-world calibration platform later.This will allow us to compare our software model and refine it.We chose Ansys Maxwell as our modeling software due to its powerful electrostatic capabilities.We create a simplified model of the Leo Rover, which only has the rough outside geometry, and contained no internal structure as this is not relevant.The boundary conditions were set to represent a typical atmospheric electric field magnitude of 100 V/m with the rover "sitting" on the ground.The upward-facing field mill aperture is located 27 cm above the ground, and hence there is an undisturbed electric field expected of 27 V/m.The results of the electrostatic modelling, shown in Figure 3a and Figure 3b, suggest that the field enhancement is minimal, because the modelled reading is close to 27 V in the horizontal plane of the aperture (Figure 3b).

Real-World Trials
We have started doing real-world experiments using one Leo Rover at this point.These initial trials indicate the sort of data that could be collected by one robot, which could be multiplied in a multi-robot, sparse swarm configuration.

Trial Methodology
Our initial experiments are relatively simple.We put the Leo Rover in a large open space (Waterfront Square, Bristol Harbourside, UK) whereby we could get a strong GNSS signal.The rover was then teleoperated to move in a square path while recording data from the electrostatic sensor.The RealSense camera was put in an upward-facing orientation using the pan-tilt mechanism to record a view of the sky, to see how many clouds or obstructions were above the robot.The rover moved in an anti-clockwise motion, starting at the location marked in Figs. 4, 5.
This process was repeated in two trials that were as similar as possible with teleoperation.The rover was kept stationary for around 10 seconds before moving in order to set a baseline reading.It is important to note that the field mill measures the electric field, not the potential gradient (electric field multiplied by −1) which is often used in atmospheric electricity research.We therefore expect a negative atmospheric electric field to be recorded by the field mill during "fair weather", typically defined as a fine day with less than 3/8 cloud cover, light winds and no dust, rain or fog [22].

Results
The results of the experiment can be seen in Figure 4 and Figure 5.Each trial shows a map indicating the GNSS position of the robot and the field mill reading at that location.The plot below each map shows the electrostatic reading over time, with each trial planned to last up to 180 seconds.In both trials, the rover comes back to the start before this time is reached.The data collection also starts approximately 10 seconds before the rover starts moving, providing a clear picture of the baseline reading and any changes in the field while the robot moves.

Discussion
The data from these trials provide a few insights into the structure of our system.First, the incoming data appears to be quite noisy.When the rover is back to its starting location, the electrostatic reading does not match the initial data readings.The cause of this is unclear, but  it could be due to the rover picking up charges while it moves.This issue shows the need to investigate further and get improvements to our grounding system.Second, the readings seem to increase when the rover moves across the top of the square in both trials.From viewing the footage from the upward-facing camera, we found that the rover moves under some trees which might have changed the electrostatic readings.Third, the readings at the bottom left location, which was the most open part of the route, were consistently around −90 V/m in both trials.This consistency, and the agreement with the fair weather atmospheric electric field expectation, shows that this part of the route is providing correct and reliable readings, which is an encouraging indication for our system.These findings provide good insight to guide our ongoing research.

Future Plans
Our research using mobile terrestrial platforms like the Leo Rover for electrostatic measurements is in its early stages, with several opportunities for improvement.

Calibration & Grounding
The initial trials have shown the need for a better understanding of grounding issues and how to solve them.We are planning to create a real-world calibration rig, which will let us check our software modelling results and compare them together to make sure our setup is reading the electrostatic field correctly.We are also looking into a system to remove charges that could be picked up by the wheels, potentially using conductive tyres or brushes.Additionally, we will also consider if adding a larger ground plane to the mill will help provide clearer results.

Rover Trials
Our aim is to change and improve our experiments to gather more reliable and meaningful results.The future trials will involve the rover moving autonomously between GNSS points, following a consistent path and velocity.This will allow us to compare data from different trials more effectively.We also plan to have the rover gather data on more challenging terrains.This will test the rover's performance in different conditions and allow us to see its suitability for extraterrestrial exploration.Finally, we aim to use the data gathered from the rover to create a map of the electrostatic environment with an additional focus on electrostatic anomalies or 'hotspots'.This could involve the rover exploring areas based on information gain [23], rather than following a pre-determined path.

Robot Swarms
We plan to expand the system to include multiple rovers, creating a sparse swarm control architecture.Before implementing this in the physical world, we will use a simulatorbased Agent-Based Model (ABM) to test different swarm control algorithms, with simplified communication, path planning, and exploration representations.Once this is done in the simulator we can implement them in real-world rovers.This will be done by using more Leo Rovers with the same hardware and software setups, or possibly, expand into including aerial robots by incorporating drones into a 'heterogeneous swarm' (e.g.[24], [25]).This could unlock further opportunities for electrostatic characterisation of different environments, whereby drones could be 'spotters' for targets of higher-resolution readings by ground-based robots.

Conclusion
In this paper, we have presented our initial work in using a terrestrial robot as a platform for electrostatic measurements, finding reasonably repeatable measurements of the atmospheric electric field in fair weather conditions [22].This is a novel approach with potential for environmental science and space exploration.We have explained our hardware and software setups, which builds on the Leo Rover platform and develop custom ROS packages for electrostatic measurement.Our first trials have given us promising results, successfully measuring the fair weather atmospheric electric field on a moving robot, with readings consistent with those expected on a fair weather day.Our results also highlight potential issues ahead, from robot grounding to calibration.Nevertheless, as we continue to improve our system and expand it into a multi-robot swarm, we are positive about the potential of this novel approach.By combining terrestrial robotics with electrostatics, we are creating new possibilities for exploration and discovery, both on Earth and in space.

Figure 1 :
Figure 1: Leo Rover (left: back view and right: front view) with labelled sensors, including GNSS, stereo image camera, 2D LiDAR, 4G antenna for long-range cellular communication and the electric field mill.Copper tape is placed to create a common ground, including attaching a grounding chain.

Figure 2 :
Figure 2: Schematic of level shift circuit to convert bipolar voltages from the field mill into a unipolar signal for input into a 12-bit ADC for logging (a) Side profile of field around rover.A black 'X' marks the location where the upward-facing field mill reads the electric field.The robot contacts the ground at 0 V, although owing to the low resolution colour banding it appears negative.(b) Top view of electric field.The horizontal plane is at the field mill reading point.The depth camera, pan-tilt base and 4G antenna unit is visible.

Figure 4 :
Figure 4: Trial 1, Date: 21/07/2023, Start time: 09:24 AM UTC, Scattered cumulus clouds.Top: Red X marks starting point and the arrow marks the direction of movement.Bottom: the electrostatic field mill reading, where the dotted red line marks the start of movement.

Figure 5 :
Figure 5: Trial 2, Date: 21/07/2023, Start time: 09:30 AM UTC, Scattered cumulus clouds.Top: Red X marks starting point and the arrow marks the direction of movement.Bottom: the electrostatic field mill reading, where the dotted red line marks the start of movement.