How swarm intelligence can extend the potential of research in maritime environments.
8 March, 2021
The global Ocean is the backbone for Earth as it provides support for life forms both inside the water and on the land. In fact, life on Earth began in the Ocean itself. However, more than 80% of this vast, underwater realm remains unmapped, unobserved, and unexplored by mankind , thus holding a lot of mysteries yet to be unlocked inside its womb. In other words, the Ocean needs to be explored further to understand it well.
A lot of technological advancements over the decades have helped scientists and the common man understand the beauty and essence of the Ocean, but only to some extent. For example, consider a scenario of toxic waste leaking from barrels on the sea bed , which could produce a very weak and irregular toxin gradient that is difficult for a single autonomous underwater vehicle (AUV) to traverse and locate the source (Read more about autonomous marine vehicles here). This is one such repository of scenarios that necessities advanced technologies. In this article, we talk about how the communication within a swarm of AUVs can be enhanced for efficient routing and target-finding using swarming algorithms. We also detail one bio-inspired approach towards swarming.
What are Swarm AUVs?
A swarm denotes a group of individual units that interact with each other, share and accomplish the task together as a group. The manifestation of complex group-level behaviours from relatively simple components is denoted by the term ‘swarm intelligence’ , and its physical manifestation in autonomous robots is referred to as ‘swarm robotics’ . Being an autonomous member of a swarm requires special abilities from an AUV, namely, it has to be aware of its own state and the state of its swarm . Generally, for swarm intelligence, one of the most challenging tasks is to establish adaptive cooperative communication between the swarms.
In this article, we explore how the use of underwater acoustic sensor networks for communication between swarm AUVs has the potential to make them effective, and discuss multi-hop communication combined with a bio-inspired algorithm for enhanced communication and flawless swarming.
Why Swarm AUVs?
AUVs are now being used for a variety of tasks, including oceanographic surveys, demining, and bathymetric data collection in marine and riverine environments . It makes sense to load one or two sensors on many little robots instead of loading them all on to one single, very expensive AUV . Swarm AUVs are crucial for both military and monitoring applications.
(i) Anti-Submarine Operations :
An experienced anti-submarine warfare officer searching for enemy submarines might order a shoal of small robots to be sent out by an autonomous surface vehicle (ASV) or launched from an unmanned aerial vehicle to map the ocean thermoclines well ahead of the fleet. These zones of abrupt temperature change are important in submarine operations because sonar is used to locate enemy submarine vessels, but the active acoustic signal will bounce off the thermocline due to the differential in water density.
With their fast-mapping capability over a wide area, a shoal of robots could report the exact location of the thermoclines to the anti-submarine vessel, pinpointing areas that submarines are likely to choose for hiding. Once detected, a submarine can then be tracked by follower robots. Conversely, a submarine commander could launch a shoal of micro-AUVs from their torpedo tube to locate an effective hiding place.
(ii) Anti-Drone countermeasures :
Singapore is currently testing an array of acoustic nodes that are anchored to the seabed around the island nation and will serve as a subsurface listening net for drones and submarines.
Meanwhile, the Norwegian Defence Research Establishment (FFI) has said that it has developed an acoustic array that can be towed behind a drone. This could be programmed to mimic a submarine and fool enemy submarine-detecting drones, luring them away.
Although the Ocean covers two-thirds of the surface of the Earth, it is surprisingly vulnerable to human influences such as overfishing, pollution from run-off and dumping of waste from human activity . Toxins from pesticides, fertilizers and other chemicals used on farms contaminate nearby rivers that flow into the ocean, which can cause extensive loss of marine life in bays and estuaries leading to the creation of dead zones . The most toxic waste material dumped into the Ocean includes dredged material, industrial waste, sewage sludge and radioactive waste. Dredging contributes about 80% of all waste dumped into the Ocean, adding up to several million tons of material dumped each year . These examples are just a few examples and are not limited which justifies the need for swarm AUVs or micro AUVs in oceanic environments.
Multi-hop communication in Swarm AUVs:
Underwater acoustic sensor networks (UW-ASNs) have recently attracted scientists in many oceanographic applications, i.e., pollution monitoring and disaster prevention. However, underwater networks face some critical challenges such as severely limited available bandwidth, high bit error rates, limited battery power, multi-path fading, and thus requires more research . From the networking perspective, data transmitted in a multi-hop manner aims to achieve the energy efficiency of the UWASNs since the sensors benefit from lower transmission power in a shorter distance of each hop . Besides, the selection of next-hop nodes (routing relay) in routing protocols can be optimized such that the optimal network performance (i.e., shortest end-to-end delay, minimum energy consumption or highest packet delivery ratio) can be obtained .
1. A multi-hop mechanism in Swarm AUVs:
The sensors deployed randomly in an underwater region frequently sense the surrounding environment and then transmit the sensed data to the destination in a multi-hop fashion. As combined with the cooperative transmission scheme, data transmission from sources to the destination involves both types of nodes: relay nodes of routing paths (routing relays) and relay nodes of one-hop cooperative communication (cooperative relays) . Generally, cooperative routing involves the following steps: neighbour table updating, RTS/CTS message exchange, relay selection, data transmission and acknowledgment. Since the ocean environment is dynamic and uncertain, whatever parameters we have chosen initially changes from time to time and it is difficult to track and maintain proper communication between any nodes. As the swarm AUVs are always in motion, communication and awareness within the swarm must always be established. For enhanced routing capabilities, we are choosing one of the phases of cooperative routing which is neighbour table updating and connecting it with a bio-inspired algorithm which is the “African Wild Dog Optimization Algorithm.”
2. Neighbour table updating:
For any wireless communication system, there are three major parameters to be considered: 1) SNR (Signal to noise ratio), 2) ToA (Time of Arrival), and 3) HC (Hop Count) . The higher the SNR, the better is the signal reception. The lower the ToA, the nearer the destination is. Lastly - the smaller the HC, the faster the signal reaches the destination. HC also indicates the presence or absence of nodes.
As shown in figure 2, every swarm AUV consists of corresponding tables, and each table consists of the required parameters to be updated frequently. The source and destination could be any swam. Now, we have the three parameters - SNR (determines how strong the links are between their neighbours), ToA (physical distances between the SM’s), and HC (useful in finding the shortest path between the source and destination). So far so good, but we have one missing piece of information that is crucial for swarm application is “Sensing.” The main aim of these swarm AUVs is to sense the gradient parameters on the surface or under the oceans that may be temperatures (thermoclines), Dissolved oxygen content, oil spills, toxic chemical accumulation, etc., and ability of the swarm AUVs to reach these accumulation sites cooperatively. This requires the swarm AUVs to follow each other based on the sensing parameter of each robot.
African Wild Dog Optimization Algorithm (AWDO):
African Wild dog packs (also referred to as Lycaon pictus) are one of the most efficient hunters amongst the canines . The percentage of successful hunts by them is nearly 80, while the success rate of lions is only 30% . African wild dogs mainly live in dry grasslands and semi-deserts in Africa, active in grasslands, savanna and open dry scrub. They live in packs and occupy territories ranging from 200 to 2000 square kilometers. African wild dogs hunt medium-sized ungulates in cooperative packs, locating by vocalizing. They can run long distances, at speeds up to about 45 kilometers per hour . Every morning the adult wild dogs group up and go hunting (figure 3). During hunts, they use high-pitched squeaks to indicate directions, and tirelessly chase their prey for 3 to 5 km, until it is eventually exhausted.
Alpha dogs are the dominant male and female; one of them is usually more dominant. The dominant alpha is in charge of making decisions such as the pack territory, the den location, the hunting time and directions. The other pack members usually follow the alpha dogs and avoid conflict. Unlike other canine species, there is a strict ranking system, the rank being determined by posturing, and they do not act aggressively—they do not even fight over food. Instead of fighting, the hunter dogs will bring a share of the meat to the sick and injured dogs. In some cases, a non-alpha dog will temporarily lead the dogs to the prey. If one of the alpha dogs dies, then the oldest member will gain the alpha status over the others. Alpha dogs use smells to make decisions and to lead the pack .
Generally, this algorithm consists of three major steps :
1- Pack Initialization: Randomly initialize the wild dog pack.
2- The alpha decision: each time, the dominant alpha dog smells several locations, evaluates them, adopts the best one and adjusts the direction.
3- The pack decision: the other dogs follow the dominant alpha dog based on fitness value. In some cases, a non-alpha dog temporarily leads the dogs or has the alpha status over the others.
4- Hoo call: any dog can call the other members with a bell-like "Hoo" sound to escape from local optima.
Three self-adaptive parameters of the pack play an important role in controlling the alpha’s movement. These parameters include the smell strength - if the smell becomes stronger, then the search area and the alpha dog’s movement becomes smaller and the target closer, and can be likened to the SNR parameter used by us. The advantage of this algorithm is that it only requires two parameters , unlike other bio-inspired algorithms that employ various initialization parameters. This lessens the computing resources at swarm AUVs. The two parameters are the number of dogs and the number of iterations. This algorithm can help with the coordination of a swarm of AUVs, via the following steps :
Step 1: InitializationRandomly initialize the pack of swarm AUVs with a neighbour table with an additional parameter to be updated as shown in figure 4.
Step2: Alpha Decision (Fitness Value)
Based on the fitness value, the swarm AUVs in the pack decide to follow the alpha who has better fitness compared to them. The fitness function is mapped to all four neighbour table parameters as shown in figure 5.
Step 3: Pack Decision
Using the fitness function values, the alpha dog is selected. If the alpha dog doesn’t have the required fitness value, then another dog in the pack with better fitness value is assigned an alpha position. This is one of the effective parts of this algorithm which can take the entire swarm AUVs to the destination as shown in figure 6.
Step 4: Hoo Call
If there is no enhancement after a few iterations, a ‘Hoo call’ is initiated which re-randomizes the positions of dogs in the pack, adds a perturbation to the last best alpha location, and calls the pack by implementing step 2 until a new, better location is found .
In summary, the potential use of a swarm intelligence algorithm for a swarm of micro AUVs in maritime applications has been described. One such algorithm based on multi-hop communication combined with a bio-inspired ‘African wild dog’ approach has been described which can help amplify cooperative awareness within the swarm. By implementing swarming technologies such as this, sensing and analysis of the vast oceans can be improved using a swarm of low-cost AUV systems.
-  NOAA, https://oceanservice.noaa.gov/facts/exploration.html accessed on June 11.
-  Schmickl, Thomas, et al. "CoCoRo--The Self-Aware Underwater Swarm." 2011 Fifth IEEE Conference on Self-Adaptive and Self-Organizing Systems Workshops. IEEE, 2011.
-  M. M. Millonas, “Swarms, phase transitions and collective intelligence,” in Artificial Life III, C. G. Langton, Ed. Reading, MA: Addison-Wesley, 1994.
-  Beni, Gerardo. "From swarm intelligence to swarm robotics." International Workshop on Swarm Robotics. Springer, Berlin, Heidelberg, 2004.
-  Paull, Liam, et al. "AUV navigation and localization: A review." IEEE Journal of Oceanic Engineering 39.1 (2013): 131-149.
-  Engineering and Technology (E&T) https://eandt.theiet.org/content/articles/2018/11/robots-at-sea-underwater-uavs-making-waves/ accessed June 12
-  Marine Bio, https://marinebio.org/conservation/ocean-dumping/ accessed June 12.
-  Tran-Dang, Hoa and Dong-Seong Kim. "Channel-aware cooperative routing in underwater acoustic sensor networks." Journal of Communications and Networks 21.1 (2019): 33-44.
-  Han, Guangjie, et al. "Routing protocols for underwater wireless sensor networks." IEEE Communications Magazine 53.11 (2015): 72-78.
-  Al Daoud, Essam, Rafat Alshorman and Feras Hanandeh. "A New Efficient Meta-Heuristic Optimization Algorithm Inspired by Wild Dog Packs." International Journal of Hybrid Information Technology 7.6 (2014): 83-100.
-  Jianqiang, Zhao, Chen Bike and Guo Xiaoli. "The Binary Code African Wild Dog Algorithm for Solving 0-1 Programming Problems in the Field of Geotechnical Engineering."
-  Creel, Scott and Nancy Marusha Creel. "Communal hunting and pack size in African wild dogs, Lycaon pictus." Animal Behaviour 50.5 (1995): 1325-1339.
-  Subramanian, C., A. S. S. Sekar and K. Subramanian. "A new engineering optimization method: African wild dog algorithm." International Journal of Soft Computing 8.3 (2013): 163-170.
About the author:
Sandeep Battula: Nature explorer and Co-Founder of Qualivon Technologies Private Limited where my intention is to connect people back to nature while adhering to the UN Nations "Sustainable Development Goals" by incorporating advanced technologies.