25/08/2023
Decision T0270/20 relates to an appeal against the (EPO) Examining Division’s decision to refuse European patent application 17382762.7 directed to the calculation and execution of an aircraft flight plan using machine learning techniques. The Examining Division had taken a predominantly negative view of the application, arguing that the subject matter of the claims was not technical, and therefore not inventive (lack of Inventive Step under Article 56 EPC), or if technical, not adequately disclosed (insufficiency under Article 83 EPC). The Appeal Board took the opposite view, arguing that the Examining Division had erred and ordering the case to be remitted back for further examination.
The case is interesting as it further demonstrates the EPO approach to the assessment of the technical character of mixed claims (claims involving non-technical features in combination with technical features), under the COMVIK decision (T0641/00 COMVIK), and highlights how the applicant’s prospects of a patent grant can be damaged by the purely subjective assessment of the Examining Division regarding what is and what is not technical, even before inventive step is properly considered. This problem is discussed further below.
The decision also discusses aspects of the sufficiency of the disclosure related to computer implemented inventions and inventions that depend on machine learning, and so provides some guidance to applicant’s concerning how much detail to disclose in patent applications that depend on machine learning techniques. This aspect of the decision is also discussed in more detail below.
Subject Matter of EP17382762.7
The invention discussed in the application related to the automatic update and execution of an airplane flight plan based on the runway configuration at a departure or destination airport, taking into account factors such as the known lay out of the airport, the wind direction, and the location of other aircraft. The value in updating the flight plan in this way was that control of the airplane could be performed more efficiently. The application noted that not all airports made runway configuration information available and known methods for estimation the configuration could be computationally expensive.
The claimed invention addressed these problems by employing an algorithm to count the number of planes in the respective cells of a 3D grid or mesh centred on the airport, determining a current configuration for the airport, and updating and autonomously executing the flight plan using a system employed on the aircraft. The algorithm required input information such as the industry surveillance data available at the airport, and the known airport layout. The full claim is set out below:
A computer-implemented method for updating an aircraft flight plan, comprising:
retrieving (210) recorded surveillance data (214) including instances of aircraft positions (216) at an airport;
determining (220) a plurality of three-dimensional surveillance cells (302) at each end (310a, 310b) of at least one runway (102) of the airport;
computing (230) a count of a number of aircraft positions (216) within each surveillance cell (302);
determining (240) a current configuration for each runway (102) based on the count computed for the surveillance cells (302) of the runway (102);
updating the flight plan of an aircraft based on the current runway configuration of the airport;
autonomously executing, by a system installed onboard the aircraft, the updated flight plan (1330).
The application was written more at a conceptual level rather than at a technological implementation level, and so focussed on the high-level nature and purpose of the problem being solved, rather than presenting detailed discussions of the underlying technical computer systems involved.
However, the required operation of the related computer systems was discussed in the application, including mention of the type of data received by the algorithm, and how this might be processed to give the desired output. Further, applicants are not usually required to discuss in an application all of the technical implementation details that make it possible to perform an invention, since operation of well known features can be left to the knowledge of the skilled person to perform.
Refusal By The Examining Division and Issues Arising from examination
The Examining Division regarded the pending independent claim as a mixed type claim, involving features that were inherently technical (the mention of computers, software and data) and other features that were not technical, but which were largely administrative or organisational (such as the determination of the airport runway configuration). Under the COMVIK approach, which exists to provide a framework for assessing the list of non-inventions in Article 52 of the EPC, only the technical features of the claim may count towards the assessment of inventive step, and on this basis, the Examining Division had argued that the claimed invention was a straightforward implementation of a non-technical features on a general purpose computer (and so not inventive).
The determination of the runway configuration for example was not considered by the Examining Division to be a technical problem since it relied on “purely algorithmic techniques applied to well known data”: mathematical methods are generally regarded as excluded subject matter unless they are applied to a clear technical problem. There is an argument that “counting planes” or “determining an airport runway configuration” is inherently a technical problem. However, the Examining Division had dismissed this as an example of a “resource scheduling problem” or a “travelling salesman” problem, in other words examples of an organisational problem that would typically be solved by experts (the notional skilled person) applying their expertise. In this case, the mere implementation of the expert’s expertise on a computer in the form of a program would not be sufficient for patentability.
Additionally, and based on a mix of the apparent common general knowledge of the skilled person, and the Examining Division’s negative view of the sufficiency of the application (or the overall level of technical detail), the claimed steps of “updating” and “autonomously executing” the flight plan were seen by the Examining Division as the mere “presentation of information” to the pilot, and thereby excluded, non-technical subject matter under Article 52 EPC.
In this regard, in the refusing decision the Examining Division observed that “a flight plan only contains instructions for a pilot and thus has no direct technical effect on the airplane or other technical system beyond its mere administration on a computer”. Thus, in the Examining Division’s view, the claimed invention offered a “broken chain” type obstacle to patentability. In other words, the claim did not necessarily define steps (in an unbroken chain) that resulted in a technical problem being solved, since the operation of the solution, and the effect of the claimed invention, would depend on in the intervention of the pilot to put it into effect.
Although the claimed step of “autonomously executing” would (at least to this patent attorney) imply actually carrying out the flight plan by manoeuvring the plane, the Examining Decision expressly took the opposite view:
The Examining Division is of the opinion that [the claim features] lack technical detail towards how those plans are executed autonomously. As apparent from the description the execution is performed by the flight management system (FMS) of the aircraft and isn’t related to updating flight plans. Executing flight/taxiing plans is the commonly known purpose of an FMS. Furthermore, in normal operation of the FMS, the “execution of the flight plan means to display the flight plan information to the pilot who then steers the airplane. Thus, there isn’t a technical effect on the plane beyond a display of flight plan information.
However, none of these features are actually present in claim 1. The feature of “autonomously executing” lacks technical detail such that a continuous guidance cannot be derived therefrom either. Also this lack of technical details leads to the conclusion that such an execution of a flight plan must be considered to belong to the common general knowledge of the skilled person. Otherwise, it would result in a lack of disclosure in the application documents (Article 83 EPC).
The above statements from Examining Division offer a number of options for rebuttal.
First, other decisions of the EPO have recognised that claimed data and/or software (in this case for example the flight plan) might have an inherent technical effect once used, irrespective of whether that use is explicitly claimed. This reasoning is partly why claims to “programs” are patentable at the EPO, despite “programs for a computer” as such being expressly excluded from patentability under Article 52 EPC. The point is that the purpose of the program is to be executed on computer, and when executed on the computer, the technical effect is inevitably achieved. Thus, as recently reiterated in decisions like G1/19 (see discussion here), it is not always necessary for claims to explicitly recite a step that results in the effect (though that claim might be clearer if it does), since the technical effect might be regarded as inherently present in the claim already.
Second, the idea that the pilot would not “automatically” apply the updated flight plan, but would instead merely use it as a route map or guidance (the travelling salesman objection) would not seem to be correct. There is guidance from the UK courts for example (see discussion here) that establishes that output data in the form of an alarm (even if requiring subsequent intervention by a human operator) is sufficient to give a technical effect to a claim, because the alarm not only imparts technical information about the underlying system, but would also invariably be acted upon by the operator.
In both of these examples, therefore the chain could be argued to be largely (or entirely) un-broken.
Lastly, the Examining Division seemed to interpret “autonomously executing” as meaning something other than the words taken at face value, arguing that the skilled person would know that the term could include merely “presenting the flight plan” to the pilot, and that if the term was supposed to mean more, then this would not be sufficiently supported by the description.
If this view was based on knowledge of the Examiner, it had not seemingly been supported by any prior art reference, and seems to go beyond a normal interpretation of the claim language. Arguably, the Examining Division was perhaps looking for language such as “controlling the plane by means of an autopilot function” or some similar clarification to make the point clear, but as discussed above, such an explicitly claimed step is not always required.
View of the Appeal Board and Remittal
The Appeal Board roundly rejected the Examining Division’s objections. They argued that nothing in the application suggested that the expression “autonomously executing” should be interpreted in the narrow manner followed by the Examining Division. The Board specifically did not agree with the Examining Division’s assessment of the function of the FMS (quoted above), pointing out that “the function of a FMS in modern airplanes is not limited to presenting flight plans to the pilot. It executes a number of functions which include flight planning, navigation, lateral and vertical guidance, and can be operatively connected to the autopilot, to which the calculated steering and thrust commands can be transmitted for execution”. Thus, the skilled person would understand the claim as implying a continuous and autonomous guidance of the aircraft according to the updated plan, such that Examining Division’s arguments regarding a possible “broken technical chain” must fail. Further, the Board argued that “the skilled person would be able, without undue burden, to program commonly known on-board flight control systems to autonomously execute the updated flight plan. The requirements of Article 83 EPC are therefore respected”.
Lastly, the Board argued that “all the features of claim 1 have a technical character and synergistically interact to automatically adapt the aircraft’s route to the current runway configuration. Hence, they should have all been taken into account for the assessment of inventive step”. This was quite a significant departure from the Examining Division’s assessment of the technical character of the claimed features.
In particular, the Board agreed with the appellant that “the invention does not consist in the computer-based implementation of a resource scheduling or traveling salesman algorithm. Although operational research algorithms may well be involved in the recalculation of the flight plan, they are not even part of the claimed subject matter”, and “while the assignment and update of runway configurations may be considered administrative measures, in the Board’s view determining an airport’s current runway configuration using computer means and based on surveillance data in the manner described in claim 1 has a technical character”.
The application was therefore referred back to the Examining Division for a further examination and possible search for relevant prior art.
Comments
The positive decision of the Appeal Board demonstrates the dangers during examining of prejudging the subject matter of an application to be essentially non-technical, and then constructing the reasons for refusal after the fact. Doing so inevitably leads into a number of logical traps from which it can be difficult to escape.
Specifically, in this case, in deciding that the claim was inherently not technical, and subsequently applying the COMVIK approach, the Examining Division seemed to find nothing else in the claim to sway them from their initial negative view. Each individual claim feature was regarded as essentially non-technical for a different reason, such as the feature being a mere “mathematical method”, or a “presentation of information”, or “a broken link in a chain”, or either known to the skilled person, or if not, insufficiently disclosed.
While an assessment of a lack of inventive step, can be challenged by the applicant using the three step test set out by the EPO (identify the closest prior art, identify the difference between the claimed invention and the prior art and the underlying technical problem solved by that difference, and decide whether that difference is obvious or not), there is no similar framework that allows an applicant to challenge what the EPO regard as “technical”.
Further, the last objection noted above (to inventive step/sufficiency of the claimed features) seems in turn to be based on a separate objection to the broad formulation of the claim, and a perceived lack of overall technical description. The Examining Division had objected that “the technical parts of the application remain on a merely abstract and illustrative level”, and that the “introduced claim features lack technical detail”. The implication being that, since the purpose of the claimed invention was regarded as intrinsically unpatentable, the Examining Division would only be interested in contemplating claims (with supporting description) that related to improvements in the underlying computerised flight or airport systems, and that such improvements would then need to be adequately described.
By raising this objection, the technical purpose of the claimed invention (to control the plane according to technical data collected by sensors and operated on by a machine learning algorithm) was apparently ignored. Further, it seems to over emphasise the role of the claim in providing detail of the invention. Generally speaking, it is not the purpose of the claim to provide technical details of how something works (implied as necessary in some of the objections), since that is the role of the description. It is sufficient for the claims merely to have enough technical detail for the claimed technical effect to be apparent for the purposes of assessing inventive step (in other words, the claim must claim the essential technical features of the invention, but no more).
To this attorney, the description and claims of the application, while offering a somewhat high-level level of detail were adequate to reasonably support the claimed invention. The description noted that determining the current runway configuration may comprise machine learning and previous training based on a plurality of training surveillance data sets including instances for the plurality of all available configurations for the corresponding runway. Thus, the application gave a comprehensive account of what data was being used, the purposes for which the data was being used, the use of a cellular mesh and volumes to break down and structure the calculation, as well as means to count the aircraft in the cells and determining a configuration.
Thus, in terms of sufficiency, a clear teaching and technical framework to perform the invention was provided, to which some implementation details could be left to the skilled person to complete.
Final Remarks
The decision of the Appeal Board should therefore be reassuring to applicants filing applications for computer implemented inventions relying on machine learning. In this case, the Appeal Board endorsed the applicant’s view that the Examining Division’s narrow assessment of technical character was incorrect, and found that the description (drafted to be a conceptual description of the invention supported by a detailed technical framework for implementation) was sufficient. Thus, although there may be cases, where more technical detail of the invention implementation is required, an overly detailed discussion of operation is not needed in every case.
In Board of Appeal decisions that concern excluded subject-matter, the Board of Appeal’s decision is often negative for patentability, and as a result there are often a lot of data points showing what is not patentable, and fewer data points showing what is patentable. In this case, the positive decision is helpful in confirming where the border on patentability may actually lie.
This article is for general information only. Its content is not a statement of the law on any subject and does not constitute advice. Please contact Reddie & Grose LLP for advice before taking any action in reliance on it.
