Annex 1 – General MPAI Terminology (Normative) 38

Annex 2 – Notices and Disclaimers Concerning MPAI Standards (Informative) 41

Annex 3 – The Governance of the MPAI Ecosystem (Informative) 43

Annex 4 – Datasets for CAV research. 45

Annex 5 – ETSI Technical Report 46

Annex 6 – Some CAV Communication Technologies. 47

Annex 1 – General MPAI Terminology (Normative)

The Terms used in this standard whose first letter is capital and are not already included in Table 1 are defined in Table 17.

Table 17MPAI-wide Terms


Term Definition
Access Static or slowly changing data that are required by an application such as domain knowledge data, data models, etc.
AI Framework (AIF) The environment where AIWs are executed.
AI Workflow (AIW) An organised aggregation of AIMs implementing a Use Case receiving AIM-specific Inputs and producing AIM-specific Outputs according to its Function.
AI Module (AIM) A processing element receiving AIM-specific Inputs and producing AIM-specific Outputs according to according to its Function.
Application Standard An MPAI Standard designed to enable a particular application domain.
Channel A connection between an output port of an AIM and an input port of an AIM. The term “connection” is also used as synonymous.
Communication The infrastructure that implements message passing between AIMs
Component One of the 7 AIF elements: Access, Communication, Controller, Internal Storage, Global Storage, MPAI Store, and User Agent
Conformance The attribute of an Implementation of being a correct technical Implem­entation of a Technical Specification.
Conformance Tester An entity authorised by MPAI to Test the Conformance of an Implem­entation.
Conformance Testing The normative document specifying the Means to Test the Conformance of an Implem­entation.
Conformance Testing Means Procedures, tools, data sets and/or data set characteristics to Test the Conformance of an Implem­en­tation.
Connection A channel connecting an output port of an AIM and an input port of an AIM.
Controller A Component that manages and controls the AIMs in the AIF, so that they execute in the correct order and at the time when they are needed
Data Format The standard digital representation of data.
Data Semantics The meaning of data.
Ecosystem The ensemble of the following actors: MPAI, MPAI Store, Implementers, Conformance Testers, Performance Testers and Users of MPAI-AIF Im­plem­en­tations as needed to enable an Interoperability Level.
Explainability The ability to trace the output of an Implementation back to the inputs that have produced it.
Fairness The attribute of an Implementation whose extent of applicability can be assessed by making the training set and/or network open to testing for bias and unanticipated results.
Function The operations effected by an AIW or an AIM on input data.
Global Storage A Component to store data shared by AIMs.
Internal Storage A Component to store data of the individual AIMs.
Identifier A name that uniquely identifies an Implementation.
Implementation 1.     An embodiment of the MPAI-AIF Technical Specification, or

2.     An AIW or AIM of a particular Level (1-2-3) conforming with a Use Case of an MPAI Applic­ation Standard.

Interoperability The ability to functionally replace an AIM with another AIM having the same Interoperability Level
Interoperability Level The attribute of an AIW and its AIMs to be executable in an AIF Implem­en­tati­on and to:

1.     Be proprietary (Level 1)

2.     Pass the Conformance Tes­ting (Level 2) of an Applic­ation Standard

3.     `Pass the Perform­ance Testing (Level 3) of an Applic­ation Standard.

Knowledge Base Structured and/or unstructured information made accessible to AIMs via MPAI-specified interfaces
Message A sequence of Records transported by Communication through Channels.
Normativity The set of attributes of a technology or a set of technologies specified by the applicable parts of an MPAI standard.
Performance The attribute of an Implementation of being Reliable, Robust, Fair and Replicable.
Performance Assessment The normative document specifying the procedures, the tools, the data sets and/or the data set characteristics to Assess the Grade of Performance of an Implementation.
Performance Assessment Means Procedures, tools, data sets and/or data set characteristics to Assess the Performance of an Implem­en­tation.
Performance Assessor An entity authorised by MPAI to Assess the Performance of an Implementation in a given Application domain
Profile A particular subset of the technologies used in MPAI-AIF or an AIW of an Application Standard and, where applicable, the classes, other subsets, options and parameters relevant to that subset.
Record A data structure with a specified structure
Reference Model The AIMs and theirs Connections in an AIW.
Reference Software A technically correct software implementation of a Technical Specific­ation containing source code, or source and compiled code.
Reliability The attribute of an Implementation that performs as specified by the Application Standard, profile and version the Implementation refers to, e.g., within the application scope, stated limitations, and for the period of time specified by the Implementer.
Replicability The attribute of an Implementation whose Performance, as Assessed by a Performance Assessor, can be replicated, within an agreed level, by another Performance Assessor.
Robustness The attribute of an Implementation that copes with data outside of the stated application scope with an estimated degree of confidence.
Service Provider An entrepreneur who offers an Implementation as a service (e.g., a recommendation service) to Users.
Standard The ensemble of Technical Specification, Reference Software, Confor­man­ce Testing and Performance Assessment of an MPAI application Standard.
Technical Specification (Framework) the normative specification of the AIF.

(Application) the normative specification of the set of AIWs belon­ging to an application domain along with the AIMs required to Im­plem­ent the AIWs that includes:

1.     The formats of the Input/Output data of the AIWs implementing the AIWs.

2.     The Connections of the AIMs of the AIW.

3.     The formats of the Input/Output data of the AIMs belonging to the AIW.

Testing Laboratory A laboratory accredited by MPAI to Assess the Grade of  Performance of Implementations.
Time Base The protocol specifying how Components can access timing information
Topology The set of AIM Connections of an AIW.
Use Case A particular instance of the Application domain target of an Application Standard.
User A user of an Implementation.
User Agent The Component interfacing the user with an AIF through the Controller
Version A revision or extension of a Standard or of one of its elements.
Zero Trust


Annex 2 – Notices and Disclaimers Concerning MPAI Standards (Informative)

The notices and legal disclaimers given below shall be borne in mind when downloading and using approved MPAI Standards.

In the following, “Standard” means the collection of four MPAI-approved and published documents: “Technical Specification”, “Reference Software” and “Conformance Testing” and, where applicable, “Performance Testing”.

Life cycle of MPAI Standards

MPAI Standards are developed in accordance with the MPAI Statutes. An MPAI Standard may only be developed when a Framework Licence has been adopted. MPAI Standards are developed by especially established MPAI Development Committees who operate on the basis of consensus, as specified in Annex 1 of the MPAI Statutes. While the MPAI General Assembly and the Board of Directors administer the process of the said Annex 1, MPAI does not independently evaluate, test, or verify the accuracy of any of the information or the suitability of any of the technology choices made in its Standards.

MPAI Standards may be modified at any time by corrigenda or new editions. A new edition, however, may not necessarily replace an existing MPAI standard. Visit the web page to determine the status of any given published MPAI Standard.

Comments on MPAI Standards are welcome from any interested parties, whether MPAI members or not. Comments shall mandatorily include the name and the version of the MPAI Standard and, if applicable, the specific page or line the comment applies to. Comments should be sent to the MPAI Secretariat. Comments will be reviewed by the appropriate committee for their technical relevance. However, MPAI does not provide interpretation, consulting information, or advice on MPAI Standards. Interested parties are invited to join MPAI so that they can attend the relevant Development Committees.

Coverage and Applicability of MPAI Standards

MPAI makes no warranties or representations concerning its Standards, and expressly disclaims all warranties, expressed or implied, concerning any of its Standards, including but not limited to the warranties of merchantability, fitness for a particular purpose, non-infringement etc. MPAI Standards are supplied “AS IS”.

The existence of an MPAI Standard does not imply that there are no other ways to produce and distribute products and services in the scope of the Standard. Technical progress may render the technologies included in the MPAI Standard obsolete by the time the Standard is used, especially in a field as dynamic as AI. Therefore, those looking for standards in the Data Compression by Artificial Intelligence area should carefully assess the suitability of MPAI Standards for their needs.


MPAI alerts users that practicing its Standards may infringe patents and other rights of third parties. Submitters of technologies to this standard have agreed to licence their Intellectual Property according to their respective Framework Licences.

Users of MPAI Standards should consider all applicable laws and regulations when using an MPAI Standard. The validity of Conformance Testing is strictly technical and refers to the correct implementation of the MPAI Standard. Moreover, positive Performance Assessment of an implementation applies exclusively in the context of the MPAI Governance and does not imply compliance with any regulatory requirements in the context of any jurisdiction. Therefore, it is the responsibility of the MPAI Standard implementer to observe or refer to the applicable regulatory requirements. By publishing an MPAI Standard, MPAI does not intend to promote actions that are not in compliance with applicable laws, and the Standard shall not be construed as doing so. In particular, users should evaluate MPAI Standards from the viewpoint of data privacy and data ownership in the context of their jurisdictions.

Implementers and users of MPAI Standards documents are responsible for determining and complying with all appropriate safety, security, environmental and health and all applicable laws and regulations.


MPAI draft and approved standards, whether they are in the form of documents or as web pages or otherwise, are copyrighted by MPAI under Swiss and international copyright laws. MPAI Standards are made available and may be used for a wide variety of public and private uses, e.g., implementation, use and reference, in laws and regulations and standardisation. By making these documents available for these and other uses, however, MPAI does not waive any rights in copyright to its Standards. For inquiries regarding the copyright of MPAI standards, please contact the MPAI Secretariat.

The Reference Software of an MPAI Standard is released with the MPAI Modified Berkeley Software Distribution licence. However, implementers should be aware that the Reference Software of an MPAI Standard may reference some third party software that may have a different licence.

Annex 3 – The Governance of the MPAI Ecosystem (Informative)

 Level 1 Interoperability

With reference to Figure 1, MPAI issues and maintains a standard – called MPAI-AIF – whose components are:

  1. An environment called AI Framework (AIF) running AI Workflows (AIW) composed of inter­connected AI Modules (AIM) exposing standard interfaces.
  2. A distribution system of AIW and AIM Implementation called MPAI Store from which an AIF Implementation can download AIWs and AIMs.
Implementers’ benefits Upload to the MPAI Store and have globally distributed Implementations of

–       AIFs conforming to MPAI-AIF.

–       AIWs and AIMs performing prop­rietary functions executable in AIF.

Users’ benefits Rely on Implementations that have been tested for security.
MPAI Store –       Tests the Conformance of Implementations to MPAI-AIF.

–       Verifies Implementations’ security, e.g., absence of malware.

–       Indicates unambiguously that Implementations are Level 1.

Level 2 Interoperability

In a Level 2 Implem­entation, the AIW must be an Implementation of an MPAI Use Case and the AIMs must con­form with an MPAI Applicati­on Standard.

Implementers’ benefits Upload to the MPAI Store and have globally distributed Implementations of

–       AIFs conforming to MPAI-AIF.

–       AIWs and AIMs conforming to MPAI Application Standards.

Users’ benefits –       Rely on Implementations of AIWs and AIMs whose Functions have been reviewed during standardisation.

–       Have a degree of Explainability of the AIW operation because the AIM Func­tions and the data  Formats are known.

Market’s benefits –       Open AIW and AIM markets foster competition leading to better products.

–       Competition of AIW and AIM Implementations fosters AI innovation.

MPAI Store’s role –       Tests Conformance of Implementations with the relevant MPAI Standard.

–       Verifies Implementations’ security.

–       Indicates unambiguously that Implementations are Level 2.

Level 3 Interoperability

MPAI does not generally set standards on how and with what data an AIM should be trained. This is an important differentiator that promotes competition leading to better solutions. However, the performance of an AIM is typically higher if the data used for training are in greater quantity and more in tune with the scope. Training data that have large variety and cover the spec­trum of all cases of interest in breadth and depth typically lead to Implementations of higher “quality”.

For Level 3, MPAI normatively specifies the process, the tools and the data or the characteristics of the data to be used to Assess the Grade of Performance of an AIM or an AIW.

Implementers’ benefits May claim their Implementations have passed Performance Assessment.
Users’ benefits Get assurance that the Implementation being used performs correctly, e.g., it has been properly trained.
Market’s benefits Implementations’ Performance Grades stimulate the development of more Performing AIM and AIW Implementations.
MPAI Store’s role –       Verifies the Implementations’ security

–       Indicates unambiguously that Implementations are Level 3.

The MPAI ecosystem

The following is a high-level description of the MPAI ecosystem operation applicable to fully conforming MPAI implementations:

  1. MPAI establishes and controls the not-for-profit MPAI Store (step 1).
  2. MPAI appoints Performance Assessors (step 2).
  3. MPAI publishes Standards (step 3).
  4. Implementers submit Implementations to Performance Assessors (step 4).
  5. If the Implementation Performance is acceptable, Performance Assessors inform Implementers (step 5a) and MPAI Store (step 5b).
  6. Implementers submit Implementations to the MPAI Store (step 6); The Store Tests Confor­mance and security of the Implementation.
  7. Users download Implementations (step 7).

Figure 9 – The MPAI ecosystem operation

The Ecosystem operation allows for AIW and AIF Implementations to be:

  1. Proprietary: security is verified and Conformance to MPAI-AIF Tested (Level 1).
  2. Conforming to an MPAI Application Standard: security is verified and Conformance to the relevant MPAI Application Standard Tested (Level 2).
  3. Assessed to be Reliable, Robust, Fair and Replicable (Level 3).

and have their Interoperability Level duly displayed in the MPAI Store.

Annex 4 – Datasets for CAV research


The nuScenes dataset ( is a large-scale autonomous driving dataset with 3d object annotations. It features:

  • Full sensor suite (1x LIDAR, 5x RADAR, 6x camera, IMU, GPS)
  • 1000 scenes of 20s each
  • 1,400,000 camera images
  • 390,000 lidar sweeps
  • Two diverse cities: Boston and Singapore
  • Left versus right hand traffic
  • Detailed map information
  • 4M 3D bounding boxes manually annotated for 23 object classes
  • Attributes such as visibility, activity and pose
  • New: 1.1B lidar points manually annotated for 32 classes
  • New: Explore nuScenes on SiaSearch
  • Free to use for non-commercial use

For a commercial license contact

nuImages is a large-scale autonomous driving dataset with image-level 2d annotations. It features:

  • 93k video clips of 6s each (150h of driving)
  • 93k annotated and 1.1M un-annotated images
  • Two diverse cities: Boston and Singapore
  • The same proven sensor suite as in nuScenes
  • Images mined for diversity
  • 800k annotated foreground objects with 2d bounding boxes and instance masks
  • 100k 2d semantic segmentation masks for background classes
  • Attributes such as rider, pose, activity, emergency lights and flying
  • Free to use for non-commercial use

Road Hazard data

Otonomo real-time Road Hazard data from connected passenger vehicles powers diverse road safety use cases, including mapping, accident predictions, smart cities and many more. The Otonomo Vehicle Data Platform secures, cleanses and normalizes the hazard data to make it more valuable and accessible for diverse use cases.

Annex 5 – ETSI Technical Report

ETSI specifies the Collective Perception Service (CPS) in its Technical Report [15]. The CPS includes the format and generation rules of the Collective Perception Message (CPM).

The CPM message format is (H=header, C=container, M=mandatory, O=optional).


Table 18 – ETSI Collective Perception Message format


PDU header H M protocol version, message ID and Station ID.
Management C M transmitter type (e.g., vehicle or RSU) and position.
Station Data C O transmitter heading, velocity, or acceleration etc.
Sensor Information C O
Perceived Object C O A CPM can report up to 128 detected objects
Free Space Addendum C O free space areas/volume within the sensor detection areas

Every 0.1s a CPM is generated if one of the 3 conditions is satisfied

ETSI makes use of a common coordinate system. A vehicle can communicate its absolute coordinates roll, pitch and yaw (Attitude).

Different CPM generation rules have been investigated [16].

Annex 6 – Some CAV Communication Technologies


The following categories of vehicular communication are part of the literature or industry effort:


V2V Vehicle-to-Vehicle communication between vehicles to exchange information about the speed and position of surrounding vehicles
V2I Vehicle-to-Infrastructure communication between vehicles and road infrastructure.
V2X Vehicle-to-Everything communication between a vehicle and any entity that may affect, or may be affected by, the vehicle
V2R Vehicle-to-Roadside communication between a vehicle and  Road Side Units (RSUs).
V2P Vehicle-to-Pedestrian communications between a vehicle and (multiple) pedestrian device(s) and to other vulnerable road users, e.g., cyclists, in close proximity
V2S Vehicle-to-Sensors communication between a vehicle and its sensors on board
V2D Vehicle-to-Device communication between a vehicle and any electronic device that may be connected to the vehicle itself
V2G Vehicle-to-Grid communication with the power grid to sell demand response services by either returning electricity to the grid or by throttling their charging rate
V2N Vehicle-to-Network broadcast and unicast communications between vehicles and the V2X management system and also the V2X AS (Application Server)
V2C Vehicle-to-Cloud communication with data centres and other devices connected to the internet

Technologies exist that support at least some aspects of the communication types of the table: