20 KiB
20 KiB
type:: REVIEWS tags:: year:: 2024 venue:: FRONTIERS-JOURNAL full-title:: ROSA: A Knowledge-based Solution for Robot Self-Adaptation date-start:: 16-12-2024 - 22:53 date-submitted:: external-links:: status:: DONE deadline-submission:: 01-12-2024 file:: @SUAVE: an exemplar for self-adaptive underwater vehicles parent:: todoist:: https://app.todoist.com/app/task/thank-you-for-agreeing-to-review-1531743-6WmjvJ9Vfj4fQwp8
- ### [[Highlights]]
collapsed:: true
- # Annotazioni
(16/12/2024, 22:55:42)
- “ask-and-architecture co-adaptation” ([Silva et al., 2023, p. 1](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=5&annotation=XMWMVZWB)) #a28ae5
- “reasoning uniquely about the captured knowledge at runtime 13 to decide when and how the robotic system should adapt. T” ([Silva et al., 2023, p. 1](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=5&annotation=H36SKH22)) #5fb236
- “enable robots to autonomously accomplish multiple tasks in diverse environments despite 21 internal and environmental uncertainties.” ([Silva et al., 2023, p. 1](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=5&annotation=5K228QIA)) #5fb236
- “navigation action that relies on vision-based localization cannot be executed in environments without lights, 24 but could potentially be executed with an alternative localization strategy based on lidar” ([Silva et al., 2023, p. 1](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=5&annotation=MQIAFK6M)) #a28ae5
- “y during runtime.” ([Silva et al., 2023, p. 1](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=5&annotation=GKYSMA3K)) #5fb236
- “TACA” ([Silva et al., 2023, p. 1](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=5&annotation=8GVSKXHY)) #5fb236
*Task and architecture co-adaptation.*
- “structural adaptation (i.e., activating or deactivating software components in the architecture) and parameter 30 adaptation.” ([Silva et al., 2023, p. 2](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=6&annotation=E44JY9IP)) #5fb236
- “reusability, composability, and extensibility” ([Silva et al., 2023, p. 2](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=6&annotation=48DXCHGV)) #5fb236
- “a modular architecture for self-adaptive robotic systems” ([Silva et al., 2023, p. 2](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=6&annotation=LGM6FXJH)) #ffd400
*What's the novelty with respect to existing work?*
- “a reusable knowledge model” ([Silva et al., 2023, p. 2](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=6&annotation=7ADK6ZFL)) #ffd400
*WHat are the characteristics of such a knowledge? How is it maintained? What feeds it?*
- “underwater pipeline inspection.” ([Silva et al., 2023, p. 3](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=7&annotation=FXGCTTVW)) #a28ae5
- “searching for the 80 pipeline” ([Silva et al., 2023, p. 3](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=7&annotation=48A8AZWI)) #5fb236
- “simultaneously following and inspecting the pipeline” ([Silva et al., 2023, p. 3](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=7&annotation=A6XK4T7W)) #5fb236
- “its task execution and architecture to perform A3” ([Silva et al., 2023, p. 3](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=7&annotation=24UFV6K5)) #ffd400
*Why do you need to TASK and ARCHITECTURE CO-ADAPTATION? What kind architecture adaptation is required to recharge the battery?*
- “new reaction time metric that represents the time a managing 93 system takes to react to uncertainties and adapt the managed subsystem.” ([Silva et al., 2023, p. 3](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=7&annotation=DLB7A5VN)) #5fb236
- “To the best of our knowledge, none of these works propose a solution for TACA that can be reused 99 for different applications without requiring the adaptation logic to be at least partially re-programmed 100 (see Table 1).” ([Silva et al., 2023, p. 3](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=7&annotation=ASD22IH5)) #ffd400
*It is necessary to be more clear here, maybe by referring to some explanatory examples to better discuss the claims.*
- “adapt the robot’s architecture 116 given its task plan.” ([Silva et al., 2023, p. 4](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=8&annotation=2S66H9WE)) #5fb236
- “Previous works have already proposed using knowledge models to capture how a robot is architected in 141 order to represent the available capabilities of the robot and using this knowledge at runtime for either task 142 or configuration decision-making” ([Silva et al., 2023, p. 5](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=9&annotation=2DDA6SGB)) #5fb236
- “do not capture all knowledge required for TACA” ([Silva et al., 2023, p. 5](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=9&annotation=LX2KRLBQ)) #ffd400
*What does it mean? It's vague how it's written.*
- “lack” ([Silva et al., 2023, p. 5](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=9&annotation=95JXK3XG)) #a28ae5
- “relation between the robot’s actions and capabilities” ([Silva et al., 2023, p. 5](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=9&annotation=I44AYMXZ)) #a28ae5
- “configuration alternatives” ([Silva et al., 2023, p. 5](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=9&annotation=YBJ62YKB)) #a28ae5
- “knowledge required to decide when and how the robotic system should 147 adapt” ([Silva et al., 2023, p. 5](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=9&annotation=5KMNM2QM)) #a28ae5
- “ROSA’s architecture” ([Silva et al., 2023, p. 5](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=9&annotation=AZCD9S94)) #5fb236
- “This work assumes that the robotics architecture is layered, containing a bottom functional layer, 154 an upper task decision layer, and an arbitrary number of layers in between, as common in robotics 155 architectures (Barnett et al., 2022).” ([Silva et al., 2023, p. 5](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=9&annotation=EBUT98WL)) #5fb236
- “To enable architectural adaptation, the robotic 160 architecture must be component-based, its components must be able to be activated and deactivated at 161 runtime, and its components’ parameters must be able to be adapted at runtime.” ([Silva et al., 2023, p. 5](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=9&annotation=9XGDJ9XE)) #ffd400
*Check if later there is some mention about how states are synched*
- “: (1) all knowledge required for the adaptation logic is captured in the central KB” ([Silva et al., 2023, p. 6](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=10&annotation=DF26WRIS)) #5fb236
- “no inter-component communication between the MAPE components” ([Silva et al., 2023, p. 6](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=10&annotation=YAZ5RIC3)) #a28ae5
- “he MAPE 169 components insert and read data from or to the KB via standardized interfaces” ([Silva et al., 2023, p. 6](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=10&annotation=KK3J9ULH)) #a28ae5
- “no explicit 170 coordination between the MAPE-K components.” ([Silva et al., 2023, p. 6](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=10&annotation=HBI4B264)) #a28ae5
- “Figure” ([Silva et al., 2023, p. 6](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=10&annotation=MT9AQHDU)) #ffd400
*Is it only for presentation purposes? What's the semantics of the proposed models? What is supposed to read/execute the elements that are shown in the models in Fig. 2?*
- “actions the robot can accomplish” ([Silva et al., 2023, p. 6](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=10&annotation=9NNDYWW3)) #2ea8e5
- “set of 184 functionalities the robot needs to realize an action” ([Silva et al., 2023, p. 6](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=10&annotation=FLIUWB38)) #2ea8e5
- “set of components required to realize a functionality” ([Silva et al., 2023, p. 6](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=10&annotation=NYJRJWI5)) #2ea8e5
- “possible parameters for a component” ([Silva et al., 2023, p. 6](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=10&annotation=AYQ2FGWC)) #2ea8e5
- “At runtime, ROSA performs parameter adaptation by switching the selected component 194 configurations. It performs structural adaptation by changing the selected function designs 195 and consequently the active Components. The task decision layer in combination with ROSA performs 196 TACA by selecting suitable function designs and component configurations for each Action 197 the robot needs to perform, and with the task decision layer selecting different Actions to perform 198 according to the feasible configurations.” ([Silva et al., 2023, p. 6](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=10&annotation=WA6733CL)) #ffd400
*WHat is the execution environment? Up to this point, concepts are presented at a higher level of abstractions and they are missing details conveing information on why thins actually work in practice.*
- “Architectural” ([Silva et al., 2023, p. 7](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=11&annotation=7KXDVZ8X)) #ffd400
*According to the text this should be Architectural Knowledge*
- “The adaptation heuristic knowledge enables adaptation due to changes in the environment or the 204 system’s QAs with the constraint relationship by capturing constraints on the selection of Actions, 205 Components, function designs, or component configurations in terms of measured values of 206 Measures. It enables adaptation due to component failures by capturing a Component’s status as an 207 attribute and adaptation due to changes in the robot’s task execution by capturing what Actions need 208 to be performed with the required action relationship. At runtime, adaptation is triggered when a 209 measurement violates a constraint, when a component has a failure status, or when required actions change.” ([Silva et al., 2023, p. 7](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=11&annotation=6WAQIVU8)) #ffd400
*This is high level and it does not highlight novel concepts with respect to what is existing in the control systems, the seams community.*
- “To capture the decision criteria on how to select an adaptation, the priority attribute can be used to 211 express the order of priority for selecting each function design or component configuration. 212 For more complex criteria, the estimation relationship can be used to capture the estimated impact of 213 selecting a function design, Component, or component configuration on the measured values 214 of Measures. Furthermore, a required action can relate to a Measure to indicate the preferred 215 estimation when selecting a configuration for that action. At runtime, the configuration planner 216 component exploits this knowledge to decide which configuration to select.” ([Silva et al., 2023, p. 7](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=11&annotation=ZD7HIH6N)) #ffd400
*Some concrete examples need to be shown to give the possibility to appreciate novel contributions.*
- “The proposed knowledge model was presented as a conceptual data model (CDM) in a non-machine240 readable format.” ([Silva et al., 2023, p. 10](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=14&annotation=QRA5XRC3)) #5fb236
- “Figure 3.” ([Silva et al., 2023, p. 11](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=15&annotation=KZMURWLV)) #ffd400
*EVen in this case, what is the formalism used to specify and execute the given rules?*
- “TypeDB” ([Silva et al., 2023, p. 12](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=16&annotation=HU7DLTHH)) #ffd400
*Some introductory paragraphs are need to present the technology and give preliminary notions that are needed to understand their usage later in the paper.*
- “The component-status-confi” ([Silva et al., 2023, p. 12](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=16&annotation=5CUWVXX4)) #ff6666
*Mind the format of the paper, this goes outside the page margin.*
- “ROS DiagnosticArray” ([Silva et al., 2023, p. 13](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=17&annotation=S4GK8JF9)) #ffd400
*Is there any specific format to use for created diagnostic messages?*
- “To enable ROSA to leverage ROS 2 327 mechanisms to adapt the system, the knowledge model was extended to capture knowledge about ROS 2 328 components as depicted in Figure 4.” ([Silva et al., 2023, p. 13](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=17&annotation=NEBSFQQJ)) #ffd400
*It is important to highlight/stress earlier in the paper when presenting the Component concept, that this will get instantiated when implementing the approach by using ROS technologies.*
- “PDDL problem formulation with information” ([Silva et al., 2023, p. 14](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=18&annotation=VE9CAY8T)) #5fb236
- “on whether the ROSA 355 actions are feasible or not using the aforementioned “action feasible ?action” PDDL predicate” ([Silva et al., 2023, p. 14](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=18&annotation=2VPQVPCR)) #5fb236
- “ROSA’s development” ([Silva et al., 2023, p. 16](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=20&annotation=BY2F77F4)) #ffd400
*It is important to highlight in this paper what is the novel contributions with respect to the work available at https://github.com/kas-lab/suave*
- “To solve the adaptation scenarios of the SUAVE exemplar with ROSA, the model depicted in Figure 6 410 was created.” ([Silva et al., 2023, p. 16](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=20&annotation=9PQ6NWZ7)) #ffd400
*It is necessary to clarify what is the execution environment and all the technical details that are needed to read and understand the model given in Fig. 6.*
- “ROSA model for SUAVE.” ([Silva et al., 2023, p. 16](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=20&annotation=9URS4AYZ)) #ffd400
*What is the process to create ROSA models like the one shown in Fig. 6?*
- “Critical battery level (U3), occurring only in the extended SUAVE use case, was solved with TACA by 426 extending the knowledge model with a recharge action, and a battery level constraint to the search 427 pipeline and inspect pipeline actions, representing the minimum battery level at which they can be selected. 428 Runtime behavior: when the battery level drops below 0.25, the status of both the search pipeline and 429 inspect pipeline actions is set to unfeasible, and the task decision layer cannot select them anymore. 430 Therefore, the task decision layer selects the recharge action which also triggers structural adaptation” ([Silva et al., 2023, p. 17](zotero://select/library/items/833GKVPN)) ([pdf](zotero://open-pdf/library/items/JLRJW6PI?page=21&annotation=MPTEZBBB)) #ffd400
*What is the advantage of this formalism with respect to what already exist? It is necessary to provide a comparison and explicitely mention the strenghts of the proposed approach.*
- ### [[Comments]]
- SUMMARY: The paper introduces ROSA, a knowledge-based solution aimed at tackling the issue of robot self-adaptation. This approach combines Task and Architecture Co-Adaptation (TACA) to enable robots to dynamically modify their architecture and task execution in response to environmental uncertainties and internal system changes. The authors propose a modular and reusable architectural framework that is supported by a knowledge base (KB), which aids in decision-making during runtime. The effectiveness of this framework is illustrated through an underwater robotic application.
- COMMENTS: The paper is about an interesting topic, even though it requires a revision to address the following issues:
- The proposed knowledge appears high-level and needs to differentiate itself from established methods in control systems and self-adaptive architectures (e.g., see SEAMS community contributions).
- By following my previous comment, the paper requires revision to better present the proposed approach's novelty compared to existing work in the self-adaptive research field.
- The description of the adaptation heuristic knowledge given at page 7 is abstract and does not highlight specific innovations over existing frameworks. What formalism is used to specify and execute the rules depicted in figures such as Fig. 2?
- The execution environment for the proposed system remains unclear. While concepts are presented at a high level, the paper needs more detail on the technical underpinnings.
- The proposed framework is not compared or discussed with respect to existing technologies or similar systems. The final paragraph of Section 3 needs to be expanded to enhance the presentation of existing approaches while clearly outlining their limitations, which the proposed approach aims to address. The sentence, "...However, the previous approaches do not capture all the knowledge required for TACA," should be significantly elaborated upon to provide a stronger motivation for this work.
- The authors need to justify their technical decisions. For example, what are the reasons for using TypeDB? What characteristics underlie the decisions made regarding the specific technologies adopted?
- The introduction of technologies like TypeDB at page 12 and ROS needs to be improved. In particular, introductory explanations are needed to help readers understand their role within the proposed framework.
- The use of PDDL for problem formulation given at page 14 requires further elaboration. How does it integrate with the knowledge model and ROSA’s runtime adaptation?
- I agree with the questions presented at the beginning of Section 7. However, it is important to address these questions by also demonstrating how existing baselines perform in relation to them. This will clearly highlight the strengths and limitations of the proposed approach.
- ### [[REVIEWS/Notes]]
- ### YELLOW CONCERNS
background-color:: yellow
- {{query (and [[ffd400]] [[frontiers-journal-1531743]] )}}
collapsed:: true
- ### ❓️Questions
- {{query (and [[question]] [[frontiers-journal-1531743]] )[[question]]}}
query-table:: true
query-properties:: [:block]