Blockchain Technology Secures Robot Swarms: A Comparison of Consensus Protocols and Their Resilience to Byzantine Robots

2 min read

Strobel, Castelló Ferrer and Dorigo compare three ways for 20 simulated robots in ARGoS to agree on the fraction of white tiles in a 2 m by 2 m arena: the linear consensus protocol (LCP), W-MSR, and a smart contract on a private Ethereum chain, with one Ethereum node per robot in a Docker container. Up to 7 Byzantine robots always report 0.0. A last experiment adds a Sybil attacker that forges a new identity every time step. Robots talk only within 50 cm, which gives an average degree of 2.4 and several disconnected clusters most of the time.

Findings

  • With no Byzantine robots, LCP was the most accurate. W-MSR and the blockchain paid a small price for their defences.
  • As Byzantine robots were added, LCP’s error rose quickly. W-MSR handled a few, but its error rose quickly beyond three. The blockchain approach stayed roughly constant.
  • One “stubborn” Byzantine robot means LCP and W-MSR can only reach consensus on its value in this graph, so the authors report consensus times for them only without Byzantine robots. The contract declares consensus on chain once the estimate changes by less than 0.02 between rounds.
  • Sending a transaction costs ether, so forging identities buys the attacker nothing. A single Sybil robot was enough to break both classical protocols.
  • Storage was modest: 148 bytes per transaction, a 6.8 MB chain after 1,000 s, and 33 MB after 24 hours with 20 robots.

Limits

  • Simulation only, with proof-of-work mining. After a warm-up of about 6 hours a block appears roughly every 15 s, and the authors advise against blockchains where fast reactions matter. They suggest splitting data into on-chain (security-relevant) and off-chain parts.
  • W-MSR ran on a graph with average degree 2.4, far below the connectivity its guarantee assumes, so the comparison says more about sparse swarms than about W-MSR’s theory (my reading).
  • The Byzantine model is a constant false report. One experiment used a fixed tile layout in which honest robots can report the same extreme values as Byzantine ones, which the authors expected to make detection harder.

Relevance

A good counterpoint to the W-MSR concept note: in a sparse, partitioned swarm, trimming neighbours’ values fails long before the theoretical bound, and scarcity (tokens) does the work instead. The physical-robot version is strobel-2023-byzantine-token-economy. See also Byzantine faults.