Phase 2: consensus

[jprachir]

• If the agent is in byzantine list, its reputation is reduced to 0.6 agent.reputation = max(0.1, agent.reputation - 0.4)
  next it is calculated in a different way:
  if agent_id in self.byzantine_agents:
weighted_score = bid_data["score"] * 0.3

--> for byzantine agent, removed penalty, modified reputation to agent.reputation - 0.7

• Agents vote based on their assessment; considers original (non-weighted) bid for voting
  agent_bid = bids.get(agent.agent_id, {"score": 0.0})["score"] # get all bids
  winner_bid = bids.get(winner, {"score": 0.0})["score"] # winner agent is based on winner = max(weighted_bids, key=weighted_bids.get)

--> changed bids to weighted_bids agent_bid = weighted_bids.get(agent.agent_id, {'score'}: 0.0)

• Byzantine agent picks voting randomly, and approving can tip the consensus vote = random.choice(["approve", "reject", "abstain"])

• If the bidding score is 0.0/ below threshold, that agent is not a candidate to allocate job and the consensus?

[jprachir]

Consensus workflow:

Step 1: Each agent computes its own bid/score for the job

Step 2: Everyone shares scores (all-to-all)

Step 3: Candidate winner is chosen (highest bid)

Step 4: Voting:

• Each agent checks “is this candidate winner acceptable compared to my own bid?” | Vote: ACK
• Sends "{Job_id, winner_agent, vote_for_winner_agent}” to every other agent

Step 5: Local tally

• Each agent independently counts received votes
• If ACK>threshold (e.g., 2/3) then that winner is committed