Belousov-Zhabotinsky liquid marbles in robot control

TL;DR

This paper demonstrates how Belousov-Zhabotinsky liquid marbles can be used to control a robot's movement through electrical potential responses to stimuli, opening new avenues for chemical reaction-based robotics.

Contribution

It introduces a novel method of robot control using liquid marbles with oscillatory chemical reactions and electrical sensing, integrating chemistry with robotics.

Findings

Liquid marbles respond to laser stimulation with electrical potential changes.

Electrical signals from the marbles can effectively control robot trajectories.

The approach enables chemical reaction-based control in robotics.

Abstract

We show how to control the movement of a wheeled robot using on-board liquid marbles made of Belousov-Zhabotinsky solution droplets coated with polyethylene powder. Two stainless steel, iridium coated electrodes were inserted in a marble and the electrical potential recorded was used to control the robot's motor. We stimulated the marble with a laser beam. It responded to the stimulation by pronounced changes in the electrical potential output. The electrical output was detected by robot. The robot was changing its trajectory in response to the stimulation. The results open new horizons for applications for oscillatory chemical reactions in robotics.

Full text

1 Introduction

Belousov-Zhabotinsky (BZ) reaction is an oscillatory chemical reaction, a model of excitable non-linear medium [10, 31, 26]. A non-stirred BZ solution exhibits interesting spatio-temporal patterns as a result of chemical excitation wave-fronts, e.g. target waves, spiral waves and localised wave-fragments. The oxidation waves have been used for image processing and computation. Wet electronics and computing circuits prototyped in BZ medium include chemical diodes [22], Boolean gates [34, 32], neuromorphic architectures [17, 15, 40, 35, 18] and associative memory [36, 37], wave-based counters [16], arithmetic circuits [7, 38, 46, 39, 19].

BZ controller for robots have been studied theoretically in the models of excitable automata lattices supplied with propulsive cilia [3]. A chemical processor to navigate a robot around obstacles in an arena has been prototyped in [1]. This processor, however, required images of the whole experimental arena to be prepared by a human operation in an off-line mode. The first real time BZ controller for a robot was designed and prototyped in [2]. In this case, a thin layer of BZ medium contained within a Petri dish was mounted onto a wheeled robot. Direction towards a source of stimulation was inferred, via an optical interface, from the 2D patterns of oxidation wave-fronts. Another example a BZ robotic controller is the closed loop control of a robotic hand with a thin layer BZ reactor [44]. The closed loop is achieved with photo-sensors placed underneath the Petri dish where the excitation of the BZ medium occurs from the movement of the robotic fingers. The way the robotic fingers react, in turn, is controlled by a micro-controller receiving data from the photo-sensors. The developed hybrid system was able to deliver highly complex behaviour by using just three sensors and three of the five fingers. Recently, the use of BZ gels was proposed to assemble millimetre-sized soft robots that exhibit photo-taxis, while not using any other kind of device to move around [9]. With the simulation of the chemical, along with the mechanical motion of the gels, their capabilities were unveiled. These worm-like gels can follow complicated routes based on different intensities of light, perform periodic movement resembling cilia and self-organise in groups.

Previously, the BZ medium has been utilised as an isolated system that needs specialised interfaces and data processing tools. We propose a hybrid system where the chemical system provides information to conventional electronics that control the movement of the robotic system through a direct electrical connection. Thus, this is a further step towards the final goal of an autonomous next generation of soft robots.

Previous prototypes of BZ controllers employed BZ in a Petri dish, which posed difficulties with manipulation and portability of the prototypes [2, 44]. Therefore, to overcome these difficulties, we decided to encapsulate ferroin-catalysed BZ solution in a liquid marble. A liquid marble (LM) is microlitre liquid droplets encapsulated in a hydrophobic powder coating [4]. This approach enables us to transfer and manipulate the BZ LM controller without wetting the underlying surface. Our scoping studies showed that the BZ media encapsulated in LMs exhibits ‘classical’ chemical excitation wave patterns, with mainly trigger waves observed [13]. The BZ media has been reported to be sensitive to illumination. Toth et al [41] experimentally demonstrated that visible light of the appropriate frequency, in their case it was a He–Ne laser with wavelength $632.8\text{\,}\mathrm{n}\mathrm{m}$, initiates oxidation in the ferroin-catalysed BZ reaction. Moreover, visible light of different wavelengths is proved to initiate or inhibit the dynamics of ferroin- or ruthenium-catalyzed BZ medium due to the photochemical properties of the catalyst [33, 43, 23, 14, 20]. We also use a laser beam to stimulate BZ LMs onboard of a Zumo robot.

The BZ LMs were mounted on and electrically interfaced with the Zumo robot [29]. The alternation in the dynamics of the reagents inside the BZ LM can be monitored potentiometrically with two iridium coated stainless steel electrode. Several paradigms of studies that use electrical potential to monitor the oscillation in a BZ system were previously published [8, 14, 28]. The robot is attractive in its simplicity of design and control. It has been used previously in studying route-following by klinokinesis, inspired by the navigation skills of desert ants [24], randomised algorithm mimicking biased lone exploration in roaches  [21], and the self optimisation procedure on a line-tracing application by using a evolutionary computing algorithm [45].

2 Methods

Belousov-Zhabotinsky (BZ) liquid marbles (LMs) were produced by coating droplets of BZ solution with ultra high density polyethylene (PE) powder (Sigma Aldrich, CAS 9002-88-4, Product Code 1002018483, particle size $100\mu m$). The BZ solution was prepared using the method reported by Field [11], omitting the surfactant Triton X. $18\text{\,}\mathrm{\textsc{m}}$ Sulphuric acid \ceH2SO4 (Fischer Scientific), sodium bromate \ceNaBrO3, malonic acid \ceCH2(COOH)2, sodium bromide \ceNaBr and $0.025\text{\,}\mathrm{\textsc{m}}$ ferroin indicator (Sigma Aldrich) were used as received. Sulphuric acid ($2\text{\,}\mathrm{ml}$) was added to deionised water ($67\text{\,}\mathrm{ml}$), to produce $0.5\text{\,}\mathrm{\textsc{m}}$ \ceH2SO4, \ceNaBrO3 was added to the acid to yield $70\text{\,}\mathrm{ml}$ of stock solution.

Stock solutions of $1\text{\,}\mathrm{\textsc{m}}$ malonic acid and $1\text{\,}\mathrm{\textsc{m}}$ NaBr were prepared by dissolving $1\text{\,}\mathrm{g}$ in $10\text{\,}\mathrm{ml}$ of deionised water. In a $50\text{\,}\mathrm{ml}$ beaker, $0.5\text{\,}\mathrm{ml}$ of $1\text{\,}\mathrm{\textsc{m}}$ malonic acid was added to $3\text{\,}\mathrm{ml}$ of the acidic \ceNaBrO3 solution. $0.25\text{\,}\mathrm{ml}$ of $1\text{\,}\mathrm{\textsc{m}}$ NaBr was then added to the beaker, which produced bromine. The solution was set aside until it was clear and colourless (ca. $3\text{\,}\mathrm{min}$) before adding $0.5\text{\,}\mathrm{ml}$ of $0.025\text{\,}\mathrm{\textsc{m}}$ ferroin indicator.

BZ LMs were prepared by pipetting a $75\mu L$ droplet of BZ solution, from a height of ca. $2\text{\,}\mathrm{mm}$ onto a powder bed of PE, using a method reported previously [13]. The BZ droplet was rolled on the powder bed for ca. $10\text{\,}\mathrm{s}$ until it was fully coated with powder.

For the initial experiments, which aimed to establish the electrical potential outputs of a BZ LM, a LM was placed in Petri dish and pierced with two iridium coated stainless steel electrodes (Fig. 1a). For experiments investigating the electrical potential of a BZ LM stimulated with a laser, sub-dermal needle electrodes with twisted cables were used (© SPES MEDICA SRL Via Buccari 21 16153 Genova, Italy). Electrical potential outputs were recorded with an ADC-24 high resolution data logger (Pico Technology, St Neots, Cambridgeshire, UK), sampling every $10\text{\,}\mathrm{ms}$.

BZ LMs were mounted on the robot by rolling the LMs into plastic holders, which were subsequently attached to the robot (Fig. 1b) and then the LMs pierced with two iridium coated stainless steel electrodes (Fig. 1c).

The robot used was a Zumo robot [29], which was an off-the-shelf solution. The robot is developed as an Arduino shield to provide a convenient interface with its controller. The algorithm that governs the trajectory of the robot is loaded on the Arduino board and the electronics necessary to power the motors are accommodated on the robot shield (Fig. 1d). Light stimulation was performed using a green laser pointer, wavelength 532$nm$, 5$mW$, for ca. 10$s$ (Fig. 1e). As previously reported  [41], the reduced form of the catalyst in a ferroin-catalyzed BZ medium, shows an absorption peak at 510$nm$. As a result, the choice of a wavelength of 532$nm$ is reasonably close to the peak to have significant impact in the dynamics of the reagents. A human operator have illuminated the BZ LM with a laser pen from a distance of approximately 20 $cm$. Using a FLIR ETS320 thermal camera with 0.06oC resolution we found that the illumination does not lead to a substantial increase in temperature in the marble (even illumination for over 30 sec causes just 0.2oC increase).

For the on-board recording an analogue-to-digital converter was used (ADS1118 Texas Instruments Incorporated). This was because the Arduino could read only positive values of an electrical potential and its resolution was limited to $4.9\text{\,}\mathrm{mV}$. As a result, negative values can be recorded and a higher resolution (down to $0.2\text{\,}\mathrm{mV}$) was achieved. The on-board recordings were saved on to an SD-card attached to the Arduino and started $3\text{\,}\mathrm{s}$ after the activation of the robot due to initialisation procedures. The robot is programmed with a simple algorithm to manipulate its moves in a constant way. However, this is not limiting its capabilities. Just for illustration reasons in the experiments executed in this study the algorithms dictates the robot to move 1.2$cm$ forward and turn to either direction at an angle of 3 degrees.

3 Results

As the oxidation wave-fronts are travelling within the BZ LM, an electrical potential that oscillates is observed in the electrodes. The dynamics of the wave-fronts and, thus, the oscillating potential are changing in response to the LM being illuminated by a laser. More specifically, one case studied was when the LM had a potential that oscillated around a negative value and was exposed to a laser beam while at the higher point of the oscillation in the positive region (Fig. 2(a)). The respond was inhibition of the oscillating output and a decrease of the oscillations’ amplitude as realised in Fig.  2(a). Another case was a sudden drop of potential with no significant changing in the oscillation characteristics (Fig. 2(b)).

Given the aforementioned observations of the effect the laser beam causes, we developed the algorithm that would navigate the robot by taking values of the potential from the BZ LM as follows. The algorithm, loaded to the Arduino board connected to the Zumo robot, reads the outputs from a BZ LM and if the value is positive then the robot turns left. Whereas, if the value read is negative the robot turns right. In order to avoid movement when the potential output of the BZ LM is too low, a condition of the absolute value being higher than 1mV was introduced. The electrical potential of the BZ LMs is read every 2 seconds and logged on an on-board SD card for further investigation.

To enhance the comprehension of the results drawn from the robot experiments, the following figures are encoded as described here. The asterisks represent a positive potential value of the BZ LM and, hence, a left turn of the robot. Respectively, the squares in the graphs represent a negative potential value read and, hence, a right turn of the robot. The circles represent a lower value than the minimum threshold that does not dictate any movement by the robot. The dashed vertical lines represent the time slots when the laser beam stimulating the BZ LM was on, and the solid vertical lines when the laser beam was off. The $x$-axis is the time in seconds and the $y$-axis is the voltage amplitude of the BZ LM in volts.

For the first experiment involving the robot, there was no stimulation with the laser beam. The potential output and the movement of the robot is depicted in Fig. 3. The potential output oscillates around zero. Thus, the robot moves either towards the right direction or towards the left direction. Given that sampling points are equally distributed between negative and positive values, the robot is moving roughly towards a given direction.

The second experiment with the robot was executed with the interaction of the BZ LM with a laser beam. As illustrated in the results from that experiment (Fig. 4) the effects of the laser beam are altering the normal oscillation (as depicted in Fig. 3) of the BZ LM. The first point of stimulation (at the 10th second) hinders the oscillation and maintains the potential values in the positive area. As a result the robot keeps moving towards a left direction. The second moment of stimulation (at the 32nd second) reactivates the oscillation around zero and, thus, forcing the robot to swing its way towards a generally straight direction. However, the two remaining stimulations with the laser does not seem to have a detectable effect on the output potential of the BZ LM.

The results of the third experiment are featured in Fig. 5. Despite the fact that all the incidents of stimulation with the laser beam have a clear effect on the oscillation and the short term amplitude of the potential, the robot moves only by turning left. The robot actually is working its way around a circle (anticlockwise), due to the fact that the potential of the BZ LM was not allowed to reach negative values, possible due to repeated initiation of oxidation wave-fronts by laser illumination.

For the final experiment the results are depicted in Fig. 6. Here, the output was initially oscillating within negative values. After the first stimulation with the laser beam, the potential output is constantly increasing and reaches positive values. As a result the robot stops moving on a clockwise direction and starts an anticlockwise turn. The oscillation is now around zero. However the second stimulation hinders the oscillation, with values of electrical potential remaining positive longer and, thus, the robot moves on an anticlockwise turn once more.

The electrodes are penetrating through the BZ LM and the plastic container onboard of the robot. Consequently, the BZ LM is not able to move freely in the plastic container. Given that the oscillation period of the potential is similar in experiments without movement (Fig. 2) and with movement (Figs. 3 to 6), the vibrations from the robot seem not to be enough to characterize the LM as a well-mixed system. As a result, the BZ LM can be considered as a distributed-parameter system with local concentration gradients.

All the electrical potential values saved on the on-board SD card of the Arduino system were congregated and investigated. The resulted data set was used to produce the histogram presented in Fig. 7. Moreover, a fitted normal distribution of the appearances of each batch of electrical potential was plotted in Fig. 7. The mean value is 0.006 and the standard deviation 0.0159. Thus, the definition of assigning left or right turns with values around zero (which is close to the mean of 0.006) provides an almost evenly distributed motion towards both sides.

4 Discussions

This work demonstrates that the BZ reaction can be directly incorporated into the electronic circuitry of a controller for a robot. Limitations imposed by earlier prototypes of liquid phase controllers, where robots were restricted to forward speeds of ca. 1cm/s and rotation speeds of ca. 1 degree/s [2], were alleviated. The additional benefits of the BZ LM system were that no optical interfaces were required to monitor the BZ LM controller and the geometries of the oxidation wave-fronts no longer need to be analysed. Hardware and software used in previous versions of the robot [2, 44] (a light placed underneath the reaction contained within a Petri dish, a serial connection to a PC and image processing algorithms) are not necessary as the BZ LMs are electrically connected to the micro-controller that delivers the trajectory of the robot. This reduction in the complexity of the controller system shows progress towards future unconventional and soft robotics.

By encapsulating the BZ solution droplets in hydrophobic powder to form LMs, made the controllers re-configurable. In principle, it would be possible to mount as many BZ marbles as desired on-board of a robot and allow the LM ensembles to process information about the local environment and potentially make decisions based on the fusions of many stimuli. The properties of LMs can be tailored for a variety of applications by altering the encapsulated liquid and / or the powder coating [6, 5, 12, 25, 27, 30]. This means LMs can be prepared to enable them to be manipulated using electrical and magnetic fields, in addition to mechanical manipulation. Thus, robotic BZ LM controllers can be reconfigured on-flight, during the robot is in motion.

It is noteworthy that the implementation proposed here is not an ideal and ready-to-use solution. This is an initial study towards the control of robots with chemical reaction-diffusion systems through an electrical connection. It is a contribution in bringing important improvement in prototyping wet robotics bearing complex dynamics. The exact behavior of the chemical system is difficult to predict and to manipulate as noticed in the results from the experiments. Consequently, the reproduction of the results and the perfect manipulation of the potential’s oscillation were not extensively analysed in the context of this study but remain as aspects of future work.

Finally, the short time of the experiments with the BZ LM mounted on the robot are not because of reagents depletion, but in order to illustrate more efficiently the reaction of the marble to laser illumination.

Acknowledgments

This research was supported by the EPSRC with grant EP/P016677/1 “Computing with Liquid Marbles”.

Supporting information

Videos and snapshots of experiments can be found at [42].