Quality Diversity (QD) is a branch of evolutionary computation that seeks high-quality and behaviorally diverse solutions to a problem. While adversarial problems are common, classical QD cannot be easily applied to them, as both fitness and behavior depend on the opposing solutions. Recently, Generational Adversarial MAP-Elites (GAME) has been proposed to coevolve both sides of an adversarial problem by alternating the execution of a multi-task QD algorithm against previous elites, called tasks. The original algorithm selects new tasks based on a behavioral criterion, which may lead to undesired dynamics due to inter-side dependencies. In addition, comparing sets of solutions cannot be done directly using classical QD metrics due to inter-side dependencies. In this paper, we propose (1) 6 metrics of adversarial quality and diversity based on an inter-variants tournament to compare the sets of solutions, ensuring a fair comparison, and (2) propose two tournament-informed task selection methods to promote higher quality and diversity at each generation. We evaluate the variants across three adversarial problems: Pong, a Cat-and-mouse game, and a Pursuers-and-evaders game. We show that the tournament-informed task selection method leads to higher adversarial quality and diversity. We hope that this work will help further advance adversarial QD.

Large Language Models (LLMs) have demonstrated remarkable performance across various tasks. Their potential to facilitate human coordination with many agents is a promising but largely under-explored area. Such capabilities would be helpful in disaster response, urban planning, and real-time strategy scenarios. In this work, we introduce (1) a real-time strategy game benchmark designed to evaluate these abilities and (2) a novel framework we term HIVE. HIVE empowers a single human to coordinate swarms of up to 2,000 agents through a natural language dialog with an LLM. We present promising results on this multi-agent benchmark, with our hybrid approach solving tasks such as coordinating agent movements, exploiting unit weaknesses, leveraging human annotations, and understanding terrain and strategic points. Our findings also highlight critical limitations of current models, including difficulties in processing spatial visual information and challenges in formulating long-term strategic plans. This work sheds light on the potential and limitations of LLMs in human-swarm coordination, paving the way for future research in this area. The HIVE project page, hive.syrkis.com, includes videos of the system in action.

Optimized routing algorithms are of most importance for Mobile Ad-hoc NETworks. The routing protocol offers a route to the destination and adapts the path when the network changes. The popular routing algorithm, Optimized Link State Routing (OLSRv2) is evaluated in realistic real time emulations using the open source EMANE platform and the open Anglova scenario. The effect of fading is analyzed to account for the mobile wireless environment. OLSR estimates the quality of each link by processing received HELLO packets. OLSR has limited performance in scenario with fading and can lead to a poor completion ratio: 67% for the 24-node Anglova company 1 scenario. The proposed solution called OLSR+PHY uses the physical layer estimate of the Signal to Interference and Noise Ratio (SINR) to drop the HELLO packets received with poor SINR. OLSR+PHY mitigates routes flapping and data packet loss. The completion ratio of acked-messages is improved to 79% while the Round Trip Time is only increased from 0.6 to 1s.

Opportunistic networks are formed by people carrying mobile devices with wireless capabilities. When in mutual transmission range (or contact), nodes of such networks use device-to-device communication to automatically exchange data, without requiring fixed infrastructure. A few apps have already proven opportunistic networking to be a useful communication paradigm. While contacts have been studied theoretically, in practice, they depend to a large extent on the characteristics of the mobile devices, typically smartphones. Up to our knowledge, no study has considered the radio characteristics of current smartphones acting as opportunistic nodes yet. In this paper, we hence investigate the WiFi radio performances of smartphones for opportunistic networking. We start by revisiting the classical link budget by adding the impact of the phone's carrier. We then perform extensive measurements to fully characterize all components of the link budget between two smartphones. We find that the Two-Ray propagation model is the best fit as a smartphone propagation model. Eventually, to assess the capacity of opportunistic networking we evaluate a simple scenario of two pedestrians crossing on a path. In this scenario, we show that a maximum range of more than 400m can be achieved and that during the resulting contact opportunity a total of 171 MB goodput can be transferred.

Fast and accurate link quality estimation is an important feature for wireless protocols such as routing, rate switching or handover. Existing signal strength based estimators tend to be fast but inaccurate while packet statistic based approaches are more accurate but require longer estimation times. We propose a new link quality estimation approach based on chip errors in symbols for direct sequence spread spectrum transceivers. The new link quality estimator is evaluated experimentally with software defined radios on IEEE 802.15.4 for different link conditions, including multi-path and mobile scenarios. We show that our chip error based link quality estimator performs more accurately than received signal strength based estimators and much faster than the packet statistic based estimators with comparable accuracy. With our approach, only a single packet, or even a fraction of a packet (e.g., only a few symbols), is necessary to obtain similar performance as state-of-the-art approaches that require at least 10 packets.

Realistic mobility models are crucial for the simulation of Delay Tolerant and Opportunistic Networks. The long standing benchmark of reproducing realistic pairwise statistics (e.g., contact and inter-contact time distributions) is today mastered by state-of-the-art models. However, mobility models should also reflect the macroscopic community structure of who meets whom. While some existing models reproduce realistic community structure - reflecting groups of nodes who work or live together - they fail in correctly capturing what happens between such communities: they are often connected by few bridging links between nodes who socialize outside of the context and location of their home communities. In a first step, we analyze the bridging behavior in mobility traces and show how it differs to that of mobility models. By analyzing the context and location of contacts, we then show that it is the social nature of bridges which makes them differ from intra-community links. Based on these insights, we propose a Hypergraph to model time-synchronized meetings of nodes from different communities as a social overlay. Applying this as an extension to two existing mobility models we show that it reproduces correct bridging behavior while keeping other features of the original models intact.

Opportunistic networks use human mobility and consequent wireless contacts between mobile devices to disseminate data in a peer-to-peer manner. Designing appropriate algorithms and protocols for such networks is challenging as it requires understanding patterns of (1) mobility (who meets whom), (2) social relations (who knows whom) and (3), communication (who communicates with whom). To date, apart from few small test setups, there are no operational opportunistic networks where measurements could reveal the complex correlation of these features of human relationships. Hence, opportunistic networking research is largely based on insights from measurements of either contacts, social networks, or communication, but not all three combined. In this paper we report an experiment called Stumbl, as a step towards collecting rich datasets comprising social, mobility and communication ties. Stumbl is a Facebook application that provides participating users with a user-friendly interface to report their daily face-to-face meetings with other Facebook friends. It also logs user interactions on Facebook (e.g. comments, wall posts, likes). This way the contact graph, social graph, and activity graphs for the same set of users could be compared and analyzed. We report here preliminary results and analyses of a first experiment we have performed.

Delay Tolerant Networks (DTN) are networks of self-organizing wireless nodes, where end-to-end connectivity is intermittent. In these networks, forwarding decisions are generally made using locally collected knowledge about node behavior (e.g., past contacts between nodes) to predict future contact opportunities. The use of complex network analysis has been recently suggested to perform this prediction task and improve the performance of DTN routing. Contacts seen in the past are aggregated to a social graph, and a variety of metrics (e.g., centrality and similarity) or algorithms (e.g., community detection) have been proposed to assess the utility of a node to deliver a content or bring it closer to the destination. In this paper, we argue that it is not so much the choice or sophistication of social metrics and algorithms that bears the most weight on performance, but rather the mapping from the mobility process generating contacts to the aggregated social graph. We first study two well-known DTN routing algorithms - SimBet and BubbleRap - that rely on such complex network analysis, and show that their performance heavily depends on how the mapping (contact aggregation) is performed. What is more, for a range of synthetic mobility models and real traces, we show that improved performances (up to a factor of 4 in terms of delivery ratio) are consistently achieved for a relatively narrow range of aggregation levels only, where the aggregated graph most closely reflects the underlying mobility structure. To this end, we propose an online algorithm that uses concepts from unsupervised learning and spectral graph theory to infer this 'correct' graph structure; this algorithm allows each node to locally identify and adjust to the optimal operating point, and achieves good performance in all scenarios considered.

The development of mobile ad hoc systems have considerably emphasized the need for a better understanding of the factors that influence the systemspsila performance, i.e., mobility patterns, radio propagation, traffic characteristics and their interrelations. Depending on the context where the system is being used-whether at work, at home, or in some means of transportation-there will be various applications that provide benefit of using the system. These different applications in turn generate traffic with very heterogeneous characteristics. In addition, the radio propagation depends on the surrounding environment and the density of communicating nodes in the system and in other systems that compete for radio resources; hence the available data rates in the system are hard to ascertain. The uncertainty regarding both the traffic and the available data rates make it hard to evaluate the performance of the system. This uncertainty is due to the lack of deployed systems. We hence believe that research in mobile ad hoc networks has to couple theory and practice in a continuous feedback loop with experimentations running full-fledged applications. We show how this methodology has proved to be beneficial in order to overcome challenges and to determine main issues for future research in the context of our PodNet project. It has also allowed us to revisit assumptions and scepticism surrounding the feasibility of mobile ad hoc networks. PodNet is an architecture, instantiated in a prototype, that is dedicated to cooperative content distribution. It enables exchange of podcasts (for any kind of multimedia content) amongst mobile devices in a peer-to-peer fashion using IEEE 802.11 in ad hoc mode. PodNet has clear advantages over traditional content distribution approaches (e.g., newspapers, 3G) in terms of public availability, practicability, capacity, resilience to failures, and jamming.