Difference between revisions of "Comparison"

From IridiaWiki
Jump to navigationJump to search
 
(34 intermediate revisions by the same user not shown)
Line 3: Line 3:
   
 
= Comparison =
 
= Comparison =
  +
  +
== Dissemination ==
  +
  +
{| class="wikitable" style="width:100%"
  +
! Feature
  +
! Classical approach
  +
! Blockchain approach
  +
|-
  +
| Number of used opinions for voting strategy
  +
| Last 2 opinions in robot's memory
  +
| Last 2 opinions in robot's blockchain
  +
|-
  +
| Time
  ⚫
| proportional to sensor readings p, using a sample from an exponential distribution with pg, g = 10
  +
| see strategies below
  +
|-
  +
| Broadcasting
  +
| During the entire phase
  +
| During the entire phase
  +
|-
  +
| Listening
  +
| Only in the last 3s of the phase [1]
  +
| Only in the last 3s of the phase
  +
|-
  +
| Peers
  +
| Send opinions to peers if distance is below 50 cm
  +
| Connect to peers (i.e., to their Ethereum process) if distance is below 50 cm (only in the last 3 s)
  +
|-
  +
| Mining
  +
| Not applicable
  +
| In the last 3s of the dissemination state
  +
|-
  +
| Decision-making strategies:
  +
| DC, DMMD, DMVD
  +
| see strategies below
  +
|}
  +
  +
[1] This is done, such that on average the robot has the same number of neighbors and no time effects are introduced into the system.
  +
  +
== Exploration ==
 
{| class="wikitable" style="width:100%"
  
{| class="wikitable" style="width:100%"
  +
! Feature
−
! Feature !! Classical approach !! Blockchain approach
 
  +
! Classical approach
  +
! Blockchain approach
 
|-
  
|-
  +
| Time
−
|| Number of used opinions for voting strategy || Last 2 opinions in robot's memory || Last 2 opinions in robot's blockchain
  
  +
| Sample from an exponential distribution with σ = 10
  +
| Sample from an exponential distribution with σ = 10
 
|-
  
|-
  +
| Peers
⚫
|| Dissemination State || Time: proportional to sensor readings \hat{p} (sample from an exponential distribution with \hat{p}g, g = 10) || Use one of the strategies described below
  
  +
| Do not send/receive messages
  +
| Not connected to any peers
 
|-
  
|-
  +
| Mining
−
|| Exploration State|| Time: Sample from an exponential distribution (sample from an exponential distribution with σ = 10)|| Time: Sample from an exponential distribution (sample from an exponential distribution with σ = 10)
  
  +
| Not applicable
−
Disconnect from peers
  
  +
| No
 
|}
  
|}
   
Line 18: Line 65:
 
=== Strategy 1: Amount of transactions ===
 
=== Strategy 1: Amount of transactions ===
 
* Send a transaction with 1 ether in each timestep of the dissemination state
  
* Send a transaction with 1 ether in each timestep of the dissemination state
  +
* Expected behavior: The stronger the opinion, the more transactions will be sent; therefore it is more likely that one of these transactions belong to the last two in the blockchain
−
* Expected behavior:
  
  +
* Disadvantage: Is this strategy close enough (amount of votes vs. time of dissemination) to the other approach?
   
 
=== Strategy 2: Direct modulation ===
 
=== Strategy 2: Direct modulation ===
 
* Send one transaction each time a robot enters the dissemination state, include amount of ether that is proportional to the quality of the opinion
  
* Send one transaction each time a robot enters the dissemination state, include amount of ether that is proportional to the quality of the opinion
  +
* Expected behavior: Should be very similar to the non-secure version; the last two votes in the blockchain
−
* Expected behavior:
  
  +
* Disadvantage: None. Should be directly comparable to the DC strategy.
   
−
=== Strategy 3: Mining time ===
+
=== Strategy 3: Dissemination time ===
−
* Make mining time proportional to the quality of the opinion
 +
* Make dissemination time proportional to the quality of the opinion (i.e., exactly the same as in the non-secure approach)
  +
* Expected behavior: The longer a robot disseminates its opinion, the more likely it is that the transaction gets included in the blockchain (more robots receive the transaction -> more hash power).
−
* Expected behavior:
  
  +
* Only listen to other opinions in the last 3s, only mine in the last 3s
  +
* Send voting transaction in the dissemination state using the JSON-RPC interface
  +
* Open questions: When to start the mining process? Does it make sense? Is it a secure approach (i.e., how robust is it to attacks)?
   
 
=== Strategy 4: Hash-puzzle ===
  
=== Strategy 4: Hash-puzzle ===

Latest revision as of 15:36, 31 January 2017

Problem

"How to transfer the sensors readings (quality of the opinion) into a secure dissemination strategy that is similar to the existing approach?"

Comparison

Dissemination

Feature Classical approach Blockchain approach
Number of used opinions for voting strategy Last 2 opinions in robot's memory Last 2 opinions in robot's blockchain
Time proportional to sensor readings p, using a sample from an exponential distribution with pg, g = 10 see strategies below
Broadcasting During the entire phase During the entire phase
Listening Only in the last 3s of the phase [1] Only in the last 3s of the phase
Peers Send opinions to peers if distance is below 50 cm Connect to peers (i.e., to their Ethereum process) if distance is below 50 cm (only in the last 3 s)
Mining Not applicable In the last 3s of the dissemination state
Decision-making strategies: DC, DMMD, DMVD see strategies below

[1] This is done, such that on average the robot has the same number of neighbors and no time effects are introduced into the system.

Exploration

Feature Classical approach Blockchain approach
Time Sample from an exponential distribution with σ = 10 Sample from an exponential distribution with σ = 10
Peers Do not send/receive messages Not connected to any peers
Mining Not applicable No

Strategies

Strategy 1: Amount of transactions

  • Send a transaction with 1 ether in each timestep of the dissemination state
  • Expected behavior: The stronger the opinion, the more transactions will be sent; therefore it is more likely that one of these transactions belong to the last two in the blockchain
  • Disadvantage: Is this strategy close enough (amount of votes vs. time of dissemination) to the other approach?

Strategy 2: Direct modulation

  • Send one transaction each time a robot enters the dissemination state, include amount of ether that is proportional to the quality of the opinion
  • Expected behavior: Should be very similar to the non-secure version; the last two votes in the blockchain
  • Disadvantage: None. Should be directly comparable to the DC strategy.

Strategy 3: Dissemination time

  • Make dissemination time proportional to the quality of the opinion (i.e., exactly the same as in the non-secure approach)
  • Expected behavior: The longer a robot disseminates its opinion, the more likely it is that the transaction gets included in the blockchain (more robots receive the transaction -> more hash power).
  • Only listen to other opinions in the last 3s, only mine in the last 3s
  • Send voting transaction in the dissemination state using the JSON-RPC interface
  • Open questions: When to start the mining process? Does it make sense? Is it a secure approach (i.e., how robust is it to attacks)?

Strategy 4: Hash-puzzle

  • Robots have to solve a (hash-based) puzzle, whose difficulty is proportional to the quality of the opinion they want to send
  • Expected behavior:

Strategy 5: Most similar

  • Only connect to neighbors in the dissemination state
  • The longer a robot is in the state, the more other robots will receive its opinion
  • Problem: When do robots mine? Only in the last x seconds (fixed)? Or for a time proportional to their opinion?
  • Expected behavior:

Alternative

  • Use an alternative approach that is not similar to Gabri's approach
  • Expected behavior:
  • Advantages: Can tailor approach to the blockchain
  • Disadvantages: Might be harder to compare the approach and show its advantages
Retrieved from "https://iridia.ulb.ac.be/w/index.php?title=Comparison&oldid=7900"